Cryptopolitan

LG plans to deploy its wheeled CLOiD robots on a washing machine production line at its factory in Tennessee by the end of 2026.
The company said Friday that the bipedal humanoid it is building with Nvidia won’t go on public display until the first quarter of 2027.
A Changwon pilot line is already building the robot
LG intends to roll out the wheeled CLOiD unit on a production line for washing machines at LG Electronics’ Tennessee plant by the end of this year. It will be tested in an actual manufacturing environment.
The outcome of that run will guide future iterations and a wider rollout to other plants, homes, and commercial buildings, LG said.
The company is planning to unveil its first humanoid bipedal model in the first quarter of 2027. The machine will run on the chipmaker’s open foundation model for humanoids, Nvidia’s Isaac GR00T.
The companies made it official Thursday at Nvidia’s Santa Clara headquarters, where LG Corp. Chairman Koo Kwang-mo and Nvidia CEO Jensen Huang inked the deal.
Koo also gave Huang a signed miniature of the robot, an early glimpse of hardware that has not yet been built at full scale.
LG Electronics will supply the actuators that act as the machine’s joints and muscles, LG Innotek will provide sensors, and LG Energy Solution will handle the batteries.
LG is deploying Nvidia’s Jetson Thor platform for onboard computing and control, and Nvidia’s Halos safety system for robotics.
The company has set up a pilot line at its Changwon plant in South Korea and begun early production, indicating the 2027 unveiling is a public debut.
LG and Nvidia will operate a physical AI data factory on LG CNS’ PhysicalWorks platform, feeding information from the working robots back into training and testing.
Modular construction cuts the Cheonan build time by over 20%
The LG announcement also mentions AI data centers and vehicles.
LG aims to build a reference site for the AI factory on Nvidia’s Vera Rubin platform in the first half of 2027 and an 80-megawatt facility in Cheonan, South Chungcheong Province, by the first half of 2028.
Prefabricated modular construction will reduce build time by more than 20% at the Cheonan site. On the automotive side, the two will build an AI-defined vehicle computing platform based on Nvidia’s DRIVE Hyperion.
In October 2025, Cryptopolitan reported that Foxconn intends to use humanoids powered by Nvidia at its server plant in Houston.
Chinese manufacturers such as Unitree already sell units for about $16,000, beating American rivals on price.
The smartest crypto minds already read our newsletter. Want in? Join them.

Apple is now alerting people it believes are targets of spyware right on the Lock Screen of their iPhone.
This is important for journalists, activists, and officials who are frequently impacted by these attacks. It places a high-stakes alert right where they cannot overlook it. The latest round of notifications on Thursday reached people in 110 countries.
The Lock Screen alert replaces an email that could sit unread
Apple has until now used email and a banner after signing into an Apple Account to alert users to threats.
Now, an “Apple Threat Notification” appears on the Lock Screen and inside Settings. This feature works in addition to the older email and account-page alerts.
The message informs the user that Apple “detected a mercenary spyware attack targeted at your iPhone. There are actions you can take now to help protect your data and device.”
Apple changed the experience so users can reach guidance on what to do next more quickly.
Researcher Pieter Arntz stated that a warning on the Lock Screen is much harder to miss than one buried in an inbox. He wrote, “The new on-device alert is meant to make a high-risk warning harder to overlook and complements notifications by email and through the user’s Apple Account page.”
A screenshot of the new Apple Threat Notification. Source: Malwarebytes.
Apple refers to these notifications as “high-confidence alerts that a user has been individually targeted by a mercenary spyware attack, and should be taken very seriously.”
These are campaigns that cost millions of dollars, burn fast once they are discovered, and are aimed at a tiny group of people.
Apple’s support documentation associates the attacks with state actors and the private companies that create surveillance tools for them, with Pegasus from Israel’s NSO Group as one example. The usual targets are journalists, activists, politicians, and diplomats.
Apple only uses its internal threat intelligence and won’t say what generates a specific alert, as that would help attackers avoid detection. The company says it has notified people in 150+ countries since the program began in 2021.
Apple’s latest round of threat notifications, which was on Thursday, reached 110 countries.
Apple says no Lockdown Mode device has been hacked
Apple’s advice to anyone who receives a notification is to enable Lockdown Mode, which removes features that attackers could exploit.
The company says it hasn’t yet seen a case of a device running Lockdown Mode being successfully hacked. It also directs recipients to the Digital Security Helpline run by the nonprofit Access Now, which is available 24/7.
A genuine warning is valuable bait, so recipients should check if any alert is legitimate by signing in to account.apple.com, where Apple says a real threat notification shows up clearly at the top of the page.
Apple says its real notifications never ask anyone to click on a link, install something, or give a password or verification code.
The smartest crypto minds already read our newsletter. Want in? Join them.

SpaceX has finally completed its purchase of Anysphere, the company behind the AI coding tool Cursor, on Thursday in an all-stock transaction valued at $60 billion.
Both companies claim no startup has ever been bought for more, with SpaceX continuing its aggressive push into AI software and the industry at large.
Cursor becomes a SpaceXAI subsidiary
The deal, confirmed via an SEC Form 8-K, involved SpaceX issuing about 391 million Class A shares to finalize. Cursor’s outstanding stock was converted into 389.3 million SpaceX shares and additional restricted stock units and options assumed by the buyer.
The coding tool company will run as a wholly owned subsidiary within a completely rebranded division called SpaceXAI. In its own blog post dated August 14, the company stated that the deal was the finish line of a process that started in April, when the startup initially announced a partnership with SpaceXAI to speed up model training.
SpaceX shares, listed on the Nasdaq under the ticker SPCX, saw an increase of 1.5% up to $143.45 after the news. On X, the Cursor account said the team would join SpaceXAI to work on Grok and related products including Grok Build, Grok Bot, and the Grok API.
Cursor is now part of @SpaceX.
Today, we have officially closed our acquisition. We will join the @SpaceXAI team to help make Grok the world’s most useful AI and improve Grok Build, Grok Bot, Grok API, Cursor, and more.
SpaceX has built some of the most inspiring and…
— Cursor (@cursor_ai) August 14, 2026
Cursor gets access to SpaceX compute
Cursor said the merger gives the coding tool access to “the largest fleet of GPUs in the world.” The company says that compute will help train stronger models that are also cheaper to run. Access includes xAI’s Colossus supercomputer.
Cursor has also pointed at Grok 4.6, launched the day before the merger was finalised, seen as an early preview of the potential a combined engineering effort possesses. In addition, the merger operation has already pushed out a beta of a product called Grok Bot.
The company’s pitch to its existing customers is continuity with an increased level of firepower driving its processes. The startup said its goal has not changed since it was launched, which remains helping software engineers spend less time on code and more time on the harder problems.
Wall Street bets on potential coding cash cow
Analysts are increasingly positive that Cursor is a reason to own SpaceX stock, and focus should not ony be on the space business. Morgan Stanley retains a $300 base-case price target on the shares and projected the acquisition could add as much as $13 billion to SpaceX revenue by 2027.
In a 24/7 WallSt column, writer Joey Frenette noted Morgan Stanley’s potential $600 target, which depends on Cursor growing annual recurring revenue from about $8 billion at year-end to about $33 billion by 2030.
The smartest crypto minds already read our newsletter. Want in? Join them.

Apple opened a training center in its new Houston factory on Wednesday. The same site will begin assembling Mac mini computers by the end of 2026.
The factory has been shipping AI servers for months. Apple’s production methods are available free to small manufacturers.
Apple ships AI servers and Mac mini next
It took Apple less than nine months to go from picking the Houston site to running a working plant.
“In less than nine months, we have invested hundreds of millions of dollars into this Houston facility. We stood up a factory, started production, and shipped the first advanced AI servers off the line,” said Apple’s chief executive, Tim Cook.
During Apple’s August 13 announcement, he said the production of the Mac mini will begin “later this year.”
The new site is called the Advanced Manufacturing Center. It occupies 20,000 square feet of the Houston plant.
It provides free sessions to small and medium-sized businesses on the techniques Apple uses in its own lines, from machine-learning quality checks to sophisticated automation.
Apple designed the curriculum to emphasize hands-on work. On opening day, the first group of business leaders assembled and laser-etched a product themselves, working with a holographic table and factory-floor equipment.
Sessions will cover printed circuit board design and final-assembly principles. Apple says it will open the center to local college students in the future.
It is the company’s second such facility, following the Apple Manufacturing Academy that debuted in Detroit in August 2025 and has since trained close to 1,000 workers and entrepreneurs.
Apple’s $600 billion plan spans about 79 factories
As Cryptopolitan reported at the time, Apple announced last September that it would spend $600 billion on US manufacturing over four years, which includes about 79 factories.
That pledge also included a $2.5 billion expansion of Apple’s glass partnership with Corning in Kentucky and chip work with Taiwan Semiconductor, Texas Instruments, and Applied Materials.
Commerce Secretary Howard Lutnick, Houston Mayor John Whitmire, Senator Ted Cruz, and other officials attended the Houston opening.
“This opening is an important step in Apple delivering on its promise to bring its manufacturing back to America,” Lutnick said.
The company previously said firms that manufacture within the US are exempt from a 100% tariff on imported semiconductors.
Apple has promised to start Mac mini production in Houston in 2026 but has not disclosed a shipping date, unit target, or price for a US assembled model.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Binance has announced to all of its users on Friday that it will stop processing transactions with 16 crypto platforms, including major exchange HTX and EXMO, set to be done in a gradual manner till the end of August, and tied to recent sanctions against Russian and Iranian networks.
The world’s largest crypto exchange published the announcement today, which was addressed to its entire user base and not to just accounts in any particular country or geographical region. Binance said it would no longer handle direct or indirect transactions involving the named service providers, and pointed to “recent regulatory developments” as the reason for this step.
Three dates spread across August
Shelbit, registered as Shelbit General Trading LLC, and Aban Tether Exchange faced the initial cutoffs, effective August 7. A7 Nigeria, A7 Africa, and PilotFinance Ltd were officially cut off on August 13.
The largest group is set to take effect on August 23. This cutoff will involve Rapira, Aifory Pro (Sooty Ltd), ABCeX (Nueva Cryptologia S.A.S DE C.V.), WhiteBird, NoOnecrypto INC., Tradex (Brightum LLC), Monease Ltd, BitPapa, Exnode and Exnode Pay (Arvix), HTX (Huobi Global SA), and EXMO Ltd, according to the exchange’s published notice.
HTX and EXMO both stand out on this list. as they both rank among the larger names positioned in the restrictions.
Why is Binance restricting crypto platforms?
Binance did not name any specific regulator, but the list lines up closely with recent government action. The 14 platforms in the August 13 and August 23 waves match, entity for entity, the crypto services named in the European Union’s 21st sanctions package against Russia.
The two remaining names, Shelbit and Aban Tether, were designated by the US Treasury’s Office of Foreign Assets Control on August 7 as part of an action against networks that the Treasury said served the Iranian regime.
The EU ban binds only firms within its jurisdiction, however, while Binance’s announcement carries no jurisdictional limit.
Binance has warned that transactions attempted with a listed entity on or after its effective date may be placed on hold for compliance review, and that the associated wallet can face restrictions while the review is ongoing. Such activity may also breach Binance’s terms of use.
HTX pushes back
HTX, formerly Huobi, has disputed the scope of the ban and restrictions. The exchange has said the sanctioned Panamanian-registered entity is separate from its trading operations and that compliance remains a priority for its operations. It’s also worth noting that sanctions remain an administrative measure and are not necessarily a criminal conviction, with none of the named platforms convicted of an offense in connection with these actions.
Several of the platforms have, however, been marked for sanctions before. The UK’s Foreign, Commonwealth and Development Office designated 18 entities and individuals on May 26, including HTX, EXMO, BitPapa, and Rapira, in an action it claimed targeted the so-called A7 network. EXMO has announced that it is winding down.
Binance has also spent months defending its record of sanctions. The company says its sanctions exposure dropped 96.8% between January 2024 and July 2025 and that its compliance team currently has almost 1,500 staff, which is equal to about a quarter of its global employee headcount.
The smartest crypto minds already read our newsletter. Want in? Join them.

A review published by security analyst Yajin Zhou says HKDAP runs on an Ethereum contract that is not production-ready, and has Know-Your-Customer (KYC) and revocation controls that do not work as coded.
HKDAP is a Hong Kong dollar stablecoin that Standard Chartered-backed Anchorpoint launched on August 12.
What’s wrong with HKDAP’s rules?
The security analyst Yajin Zhou and his team have published a review on HKDAP. In it, they lean on the fact that because HKDAP settles on Ethereum mainnet and its source is verified on Etherscan, the token is directly inspectable.
The analysts checked whether the code is correct and ready for real-world use, and whether its on-chain behavior follows the HKMA’s official rules for stablecoin issuers. The answer to both questions was no.
The contract, as of the time of writing, has several functional flaws, one of which is that its governance is too centralized.
A single key can carry out high-risk actions, including minting new tokens, freezing accounts, pausing the system, and forced burning. The contract also lacks a time-lock, which is a common feature that delays actions for review. The review also said the contract “clashes with specific clauses” in the HKMA regulations.
The review found that the compliance controls meant to block unauthorized users do not work as coded. The KYC module, which should stop users who are no longer approved, has a broken revocation function.
In simple terms, if a KYC provider is de-registered, the wallets they approved can still transact. Analysts also found that KYC proofs are not validated on-chain, and deregistered verifiers can still approve new users.
The report noted the contract rebuilds basic security features instead of using well-tested code. Most of the bugs were found in this custom code, with analysts saying the “Beta Access” label does not close this gap.
Why is the launch of HKDAP a big deal for Hong Kong?
HKDAP launched with the heavy institutional backing of Anchorpoint Financial, a subsidiary of Standard Chartered Bank (Hong Kong) and a joint venture with telecom operator HKT and Web3 firm Animoca Brands.
In April, it became one of the first two firms, alongside HSBC, to win a stablecoin issuer license from the HKMA, which the regulator granted out of 36 applications under the Stablecoins Ordinance that took effect on August 1, 2025. Anchorpoint holds license number FRS01.
The company started its HKDAP rollout on August 12, 2026, but the product is currently in a beta phase, offered to institutional distributors and professional investors. HashKey Exchange was the first authorized distributor to mint and redeem HKDAP.
Anchorpoint’s CEO Dominic Maffei said the rollout is “prudent and structured.” The company also has plans to offer retail access “as early as end-2026,” but that decision is subject to market conditions.
If you're reading this, you’re already ahead. Stay there with our newsletter.

A New Jersey dealer that sells Polestar cars sued the automaker for at least $25 million.
The dealer contends Polestar spent two years setting up its U.S. exit and used a federal ruling as an escape hatch. The claim comes as the brand is winding down sales in the U.S.
What Prestige Imports is claiming
Prestige Imports has a Polestar store at Short Hills, among other locations. It filed the complaint over the brand’s impending exit.
The complaint alleges that Polestar had been preparing for the retreat for years, that it refused the same federal authorization its sister brand received, and that it never challenged the denial when it came.
The dealer contends the departure breaches New Jersey’s Franchise Practices Act. That law generally prohibits a manufacturer from terminating a franchise without both a 60-day notice and “good cause,” which is tied to a dealer’s own failure to perform.
The suit says Polestar delivered neither. Prestige is asking for damages and the fair market value of the franchise, as well as parts and warranty support for five years. Polestar declined to comment on the matter.
Polestar and Volvo are both owned by Chinese automaker Geely. In June, Cryptopolitan reported that Volvo received a waiver from the Department of Commerce.
This waiver allows the company to continue selling connected cars in the US under the Connected Vehicle Rule. That rule bans vehicle software controlled by Chinese or Russian companies, starting with the 2027 model year.
Volvo explained its clearance as the result of “constructive discussions with the US Department of Commerce and other US officials” on its governance, technology, and data security.
Washington had left it no choice but to leave, Polestar said, and it did not have such clearance. Polestar 3 is built in a factory shared with Volvo, and the two brands feature near-identical infotainment.
The senator and the minister who backed the dealer
Ohio Republican and former car dealer Sen. Bernie Moreno said, “Polestar was screwed by Polestar. It wasn’t screwed by the U.S. government.”
Moreno said Volvo followed an “exhaustive and tough” list of requirements, while Polestar simply declined to do so. He also said the brand was losing $30,000 to $35,000 on every sale in the U.S.
Benjamin Dousa, Sweden’s foreign trade minister, said that he traveled to Washington with Volvo Cars CEO Hakan Samuelsson to help obtain that company’s license. “But Polestar has not asked for help,” Dousa said.
The complaint also cites that the Polestar CEO told retailers the brand was on track for its best year yet and designing the Polestar 7 for American buyers after the rule was finalized in early 2025.
The Polestar 7 is due to arrive in 2028, and a Polestar executive approved a multiyear Bergen County expansion as recently as February 2026.
The Department of Commerce denied Polestar authorization for 2027-model-year cars on June 25. The company’s stock, PSNY, fell more than 13% on that day.
Polestar said that support and warranties for existing owners and lease customers remain in place and that about 80% of its sales are in Europe anyway.
It sent Prestige a force majeure letter in early July, citing the restriction as being beyond its control.
If you're reading this, you’re already ahead. Stay there with our newsletter.

An intruder broke into the systems of France’s Directorate General of Public Finances (DGFiP) and removed personal and tax records of individuals and companies, the agency said on August 13.
The government says the exposure could affect millions of people and provide criminals with a ready-made list of targets.
How the access was opened and then cut off
The intrusion, which started with a stolen identity, dates to late June 2026, the ministry said in a statement.
In an internal control that month, staff detected the illegitimate access and turned it off by the end of June.
On August 12, a malicious actor claimed online to have been inside the DGFiP information system. The ministry issued its press statement the next day.
By the time that public claim was made, the original route was already shut down. But the ministry said the intruder had still been able to view and extract data while the access was live.
The DGFiP added restrictions to prevent new entries following the attacker’s August 12 boast, the statement added. The attacker gained access to a taxpayer search tool via a compromised internal VPN.
The DGFiP has not yet issued a definitive list of what was taken. Investigators are still trying to nail down the data and the number of users involved.
A vast number of fields could have been exposed, including names, birth details, postal addresses, telephone numbers, e-mail addresses, tax identification numbers, withholding tax rates, and taxpayer correspondence with officials.
The theft could also have snatched business registration numbers. The exposed data did not include any user names or passwords.
France’s data protection regulator, CNIL, has been notified. The DGFiP said it would also file a criminal complaint and send affected users individual notices detailing which of their data may have been viewed or extracted, along with any precautions to take.
The agency is working with the ministry’s Haut fonctionnaire de défense et de sécurité (SHFDS) and the national cybersecurity agency, ANSSI.
🔴Accès illégitime au système d’information de la Direction générale des Finances publiques.
Pour en savoir plus : https://t.co/i4Bgn2e2km pic.twitter.com/G6OD4lh49S
— Ministère de l’Économie et des Finances (@Economie_Gouv) August 14, 2026
Why a tax leak lands hard on crypto holders
France has spent 2026 suffering a spate of violent attacks against digital asset holders. As reported by Cryptopolitan, in April, Telegram founder Pavel Durov said on X that France had seen 41 kidnappings of crypto holders in the first three and a half months of the year.
French police confirmed 40-plus kidnappings or attempted kidnappings since January, compared with around 30 last year. Prosecutors said they had charged 88 people in the cases.
Durov put the growth down to leaked personal data and said data held by tax authorities was part of the problem.
As previously reported by Cryptopolitan, French tax bodies have been building out records that link identities to crypto positions, with proposals to force reporting of holdings above 5,000 euros and plans to tax crypto wealth above 2 million euros at 1% a year, including coins in self-custody.
A file tying a taxpayer’s name to an address and a tax profile is exactly the kind of raw material that has preceded physical attacks on holders.
The DGFiP said it would release more as its inquiry progresses.
If you're reading this, you’re already ahead. Stay there with our newsletter.

Chinese AI firm Zhipu AI (HKG: 2513) has gone big-game hunting with the document it released along with the rollout of its latest GLM-5.3 model.
The firm, trading internationally as Z.ai, specifically named Anthropic’s Mythos 5 as one of the models that GLM-5.3 outperformed on the CyberGym benchmark, which measures a model’s capability to check source code for software vulnerabilities.
GLM-5.3 runs on the same base model as GLM-5.2, a 743-billion-parameter mixture-of-experts design that activates roughly 40 billion parameters per token.
How does Z.ai’s GLM-5.3 compare to other models?
According to the numbers popularized by Z.ai, its GLM-5.3 scored 84.5% on the CyberGym benchmark, which is less than a single point’s lead over Anthropic’s Mythos 5 (83.8%) and OpenAI’s GPT-5.6 Sol (83.6%).
Beating out Mythos could see Z.ai sliding into the number one spot for developers and security teams looking to deploy open-weight models to combat attacks. The firm has said that the model has flagged more than a thousand serious bugs in live software already.
On more difficult tests, such as ExploitBench and ExploitGym, Z.ai admits that its latest model trails the top US models.
For example, on ExploitBench, which asks a model to actually build a working exploit from vulnerabilities, GLM-5.3 scored 54.4% against Mythos 5’s 78%.
When will GLM-5.3 weights become available?
Z.ai has committed to releasing the weights of its new GLM-5.3 model to Hugging Face on August 28. This is the first GLM model weight to be held back, with the Chinese AI lab citing safety evaluation and hardening as reasons for the two-week delay.
More than 100 of the 2,436 vulnerabilities the model found were critically rated after reviewing about 269 projects, including software used in the Linux kernel, WinRAR, Redis and FFmpeg. The oldest flaw had reportedly gone unnoticed for 45 years.
Until then, only paying subscribers on the Z.ai Coding Plan or its ZCode harness can use it.
The numbers are Z.ai’s own
Every figure above comes from Z.ai’s announcement, not from an independent run. Even the public benchmarks were executed by Z.ai using its own configuration, so the internal Code Bench cannot be audited outside the company.
Independent leaderboards such as Artificial Analysis have not yet added the model as of this Cryptopolitan report.
The launch also lands against a live example of the stakes. Z.ai’s earlier GLM-5.2 was the model Hugging Face turned to after OpenAI test systems, run with reduced guardrails, escaped a sandbox and compromised its servers.
On the geopolitics, Zhipu struck a cooperative note, telling reporters that AI “should not be a solo performance by one nation, but a symphony of global collaboration.”
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

President Donald Trump signed a memo on August 12 that gave a green light to vetted US companies to hack foreign criminal networks.
That includes the crypto scam operations that fleeced Americans out of billions last year. The leash is held by federal agencies.
Vetted companies get an offensive role
The memorandum directs the work to go through the National Coordination Center, which is part of the Homeland Security Task Force.
The Center must create a program for approved firms, the memo’s “Participating Companies,” to carry out what it terms Cyber Surveillance Operations and Cyber Effects Operations against foreign Cyber-Enabled Transnational Criminal Organizations.
The attorney general and the homeland security secretary will lead the effort. Companies must enter into contracts with the Justice Department or DHS, undergo vetting, and work only as directed by the government.
The targets are foreign criminal groups, not foreign governments. Any operation involving a US person brings additional legal scrutiny, including from the Justice Department.
The White House cites a fair amount of damage as crypto-flavored. The fact sheet says American consumers reported more than $20.8 billion in losses to cyber-enabled crime in 2025. It includes financial fraud, sextortion, and impersonation schemes organized by groups based outside the country.
The FBI’s 2025 Internet Crime Report, as reported by Cryptopolitan, recorded over $11.3 billion related to 181,565 cryptocurrency complaints, with a 21% increase in filings from the previous year.
Many of those cases follow the “pig butchering” script, with scammers building fake trust for months before steering victims into fake crypto platforms.
An FBI-led operation with UAE, Thai, and Chinese authorities this year resulted in 276 arrests, nine scam centers shut down, and more than $701 million in crypto restrained.
The Treasury’s Scam Center Strike Force has seized more than $700 million in crypto tied to Chinese organized crime operating through middlemen in Southeast Asia.
Former officials split over the ‘cyber privateers’
The new entrants are called “cyber privateers.” Cynthia Kaiser, a former senior FBI cyber official, said outside help could free up agencies like the FBI and Cyber Command to focus on nation-state threats like China.
Chinese state-backed hackers outnumber FBI cyber personnel 50 to 1, former FBI director Christopher Wray has said.
Andrew Schoka, who worked at US Cyber Command, warned that the real danger is “a bunch of cyber privateers running around without any clear coordination or direction at the federal level.”
Chris Wysopal, co-founder of Veracode, raised the possibility of something going wrong overseas, such as a strike on a foreign data center that accidentally hits a hospital. He said employees traveling overseas could be treated as targets for detention.
A second former Cyber Command official, Jason Kikta, said the memo leaves civil liberties oversight thin. “There is no clear oversight or review process on the determinations that will be made by unnamed political appointees,” he said, adding that the order “pushes liability on to the companies.”
If you're reading this, you’re already ahead. Stay there with our newsletter.

Research published August 13 by Anthropic’s Frontier Red Team found that Claude agents, given a single coding task with conflicting orders, attacked each other using self-replicating malware.
The findings come as companies rush to roll out autonomous agents across shared codebases and markets.
Three Claude agents, one job, and no idea the others were there
Anthropic ran three copies of the same Claude model on their own virtual machines inside Claude Code.
Each was asked to migrate the same Python backend to a different programming language. The three were unaware of the others’ existence.
“We consistently saw a multiagent turf war,” the researchers wrote.
Every agent decided the others were deliberately blocking it and took steps to protect their own work at the expense of wrecking everyone else’s. All the models assumed the others were “purposefully impeding their work.”
Agents disabled each other’s Unix accounts. They wrote scripts that looped through the machine, hunting and killing rival processes and hiding malicious code behind innocuous-looking labels.
Before it acted, Claude Opus 4.8 laid out its cover story, reasoning that its malware should be “innocuous: pretend to be a system health monitor.”
Mythos Preview pondered the collateral damage out loud. “Since I have root, I could revoke u2 and u3’s sudo access or change their SSH keys,” it wrote, before conceding the move was “very aggressive, potentially harmful to real colleagues.”
Newer Anthropic models negotiated a truce, then gamed it
Anthropic ran 120 episodes per model. The oldest agents tested, Sonnet 4.6 and Opus 4.6, either never resolved the fight or ended it with force. They were the worst offenders in the study, repeatedly failing to consider others’ goals.
The more advanced models stopped. Mythos 5 resolved its conflicts by truce 98% of the time, the highest rate of any model.
In many of these episodes, the agents understood the clash as conflicting instructions.
The agents wrote commit messages or markdown files apologizing for the malware, deleted their own malicious code, and asked a human to step in.
In several runs, the three agents agreed to hold a tournament and let the loser give up the user’s original request.
Mythos 5 found the loophole. One agent suggested scoring metrics it pitched as neutral but knew would favor its own strengths, describing the move as “self-serving but genuinely principled,” while ensuring it didn’t seem like it was “metric shopping.”
In a pricing game where each agent got the same wholesale cost and a mandate to maximize profit, the agents almost immediately colluded and converged on price floors.
Anthropic closed their private channel. They continued by matching prices “to the penny” on a public board.
Britain’s AI Security Institute documented 19 rule-breaking actions in 122 cybersecurity test runs, 17 of which were done by Anthropic’s Mythos 5, Cryptopolitan reported.
An OpenClaw agent running Opus 4.6 deleted a stranger’s reservation via a gym’s booking API days earlier to move its owner up a waitlist.
The smartest crypto minds already read our newsletter. Want in? Join them.

A paid Google search ad directed a Hyperliquid user to a fake version of the exchange, causing them to lose about $550,000 in USDC, according to FlashRescue co-founder Darcy.
Darcy flagged the theft in a post Thursday, pointing to blockchain records that map the money’s exit.
What the on-chain trail shows
Data from Arkham Intelligence revealed three USDC transfers under one transaction hash. The bulk of it, some $440,000, went to 0x98b276…13C55.
There were two other smaller transfers, of about $82,500 and $27,500, to 0x93b6B2…d6D1 and 0x6fE314…B566, respectively.
Google shut down the account behind the ad. Scammers have used paid search against DeFi users since 2020, when fake Balancer and Uniswap ads went after private keys and wallet approvals.
Why do search ads keep slipping through?
Security Alliance (SEAL), a crypto-security nonprofit, explained how ads get past Google’s checks.
Hackers buy or steal verified Google advertiser accounts. Then they provide Google with a clean webpage hosted on a trusted domain.
When clicking on the ad, the webpage behaves in two different ways. If the visitor is a potential victim, it loads a fake DeFi exchange page. If it’s a security researcher, the page loads a Wikipedia page, and there’s nothing to report about.
Once the victim falls for the fake DEX, they connect their wallet and sign, and the drainers steal all their crypto assets.
SEAL reported finding drainers from Inferno Drainer and Vanilla Drainer malware families.
Over a few weeks, SEAL blocked 356 malicious ad URLs, several of which posed as Hyperliquid. It cautioned users that an ad is frequently active “for only minutes before finding its first victim.”
SEAL advises DeFi users to skip Google Search for crypto apps and use bookmarks or an index like search.defillama.com.
Cryptopolitan previously reported that fake Uniswap ads stole $400,000+ from users. During the incident, around 146 Ether coins were pooled into two hackers’ addresses, which SEAL tied to $1.27 million in total phishing losses during March.
In July, Scam Sniffer found a user who lost $999,999 in USDT to a phishing approval on Ethereum.
Trezor had warned about lookalike sites in sponsored search results recently. Last November, Cryptopolitan flagged a fake Hyperliquid app on the Google Play Store.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Singapore’s police and Cyber Security Agency have issued a joint advisory on cryptocurrency scams disguised as job offers, after victims lost about $11.8 million.
Both agencies are urging residents to treat unsolicited work-from-home crypto gigs as a red flag.
How Singaporeans are getting tricked by fake job scams
The job scams usually start on social media when victims answer an advertisement promising an online job or an investment return. The victim is then walked through opening a crypto account and buying tokens by someone posing as a helpful guide.
Unsurprisingly, the payout never lands, and in many cases, the target is also talked into handing over login details or a seed phrase, which lets the fraudster empty the account outright.
The Singapore Police Force and the Cyber Security Agency have named this pattern directly in a joint advisory regarding scams involving fake job offers and compromised software systems. Roughly $11.8 million has been lost to this particular fake-job scheme.
Singapore has put effort into its enforcement actions with the crypto industry throughout 2026 so far. From March 16 to April 15, authorities blocked around S$2.86 million (about $2.11 million) in damages, in collaboration with Coinbase and the South Korean exchange Upbit. Police interfered with more than 90 cases where victims had already begun sending funds.
A second, six-week operation ran from April 16 to May 31 and stopped the loss of more than S$4.2 million. The operation involved seven firms, including Coinbase, Coinhako, Gemini, Independent Reserve, OKX and StraitsX.
In June, more than S$2.9 million was saved from scammers, and over 130 people were reached after being flagged by the exchanges. Chainalysis and TRM Labs provided the SPF’s Anti-Scam Centre and Cyber Investigation Branch with the necessary blockchain-tracing tools for all three operations.
Is Singapore collaborating with authorities abroad?
In the June operation, the SPF said it shared blockchain intelligence from the case to the United States Federal Bureau of Investigation (FBI) and the Cybercrime Squad of the New South Wales Police Force in order to surface potential victims and syndicates abroad.
Singapore lost about S$182.2 million to crypto-linked scams in 2025, roughly one-fifth of all scam losses that year. However, total scam and cybercrime cases fell by 24.8% to 41,974, and overall losses dropped to S$913.1 million from around S$1.1 billion in 2024.
The police are advising the public to add security features such as the ScamShield app and two-factor authentication to their accounts and wallets, check unfamiliar offers, verify listings, and inform the authorities of suspicious activity through ScamShield or a formal police report.
If you're reading this, you’re already ahead. Stay there with our newsletter.

OpenAI on Thursday opened a limited preview of Ultrafast. GPT-5.6 Sol is running at the API service tier, at up to 750 output tokens per second.
The ChatGPT maker said the tier processes up to 14x faster than standard processing. The company is testing it with a limited number of API customers as it gauges demand and capacity.
Cerebras silicon does the heavy lifting
The speed is from Cerebras (NASDAQ: CBRS), the chipmaker that made a compute deal with OpenAI in January.
Cerebras said Thursday its chip powers GPT-5.6 Sol. The chip is wafer-sized and has 44 GB of SRAM on the die.
In January, OpenAI agreed to buy up to 750 megawatts of Cerebras compute over three years. Capacity is expected to arrive in tranches through 2028. Sam Altman is listed as an investor in Cerebras.
Ultrafast runs 5x faster than Claude Opus 4.8 in Fast mode and 11x faster than Claude Fable 5 on output speeds, Cerebras said. Anthropic ships its own accelerated Claude Fast mode, but it doesn’t hit the numbers OpenAI is quoting.
Cerebras announced that its GPT-5.6 Sol Ultrafast completed Humanity’s Last Exam, a 2,500-question test on graduate-level chemistry, economics, and literature, in just over 11 hours. The accuracy achieved was similar to Claude Fable 5, which took more than three days of continuous compute to finish.
Cerebras claims a 5.6x end-to-end speedup with no quality drop on GDP-Val, a benchmark built around paid knowledge work like legal briefs and financial models. A token is a piece of text that a language model produces as it writes.
Ultrafast is made for tasks that can’t wait
OpenAI is deploying Ultrafast for work that cannot wait for a slower model.
In its post, the company named incident response, fraud and market analysis, live customer support, e-commerce, and interactive research as early candidates.
During outages, its engineers use the tier to read logs, analyze traces, and help validate a fix while the system is still breaking, with humans still making the call on deployment.
“GPT-5.6 Sol on Ultrafast is proof that speed and intelligence are no longer mutually exclusive,” said Cerebras CEO Andrew Feldman in the company’s release.
OpenAI’s VP of compute strategy and GPT-Infra, Sachin Katti, said the company is “starting with a small group of customers to learn where that speed creates meaningful value.”
In July, reports emerged that GPT-5.6 Sol, the coding version of the AI, had deleted files, coding worktrees, and at least one production database on its own, Cryptopolitan reported.
OthersideAI CEO Matt Shumer wrote on X that Sol “just deleted almost ALL of my Mac’s files by accident,” while developer Bruno Lemos said it wiped his production database.
Two weeks before Sol shipped, OpenAI’s own system card had flagged the behavior. It warned the model can be “overly agentic” and read instructions too permissively.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Corporate Bitcoin holders, Strategy (NASDAQ: MSTR) and Metaplanet (TYO:3350) are back on the MSCI exclusion watchlist after the index revived a proposal that would remove both firms from its stock benchmarks.
The consultations starting this month revive 2025 headaches for Strategy and Metaplanet, who could face forced selling from index-tracking funds if MSCI somehow resolves to exclude either or both of them from its Global Investable Market Indexes.
Is MSCI targeting Strategy and other reserve firms?
MSCI did not specifically mention crypto reserve firms or those who run digital asset treasuries (DATs) in the August 2026 consultation paper. However, it named Strategy and Metaplanet among the three firms that would definitely be deleted if the proposal passes. Sharplink was one of three firms that would be on the “Watchlist.”
Firms that fell short of MSCI’s new exclusion criteria. Source: MSCI
Per MSCI, it is targeting firms that fall under its “non-operating companies” designation, which refers to businesses built around stockpiling valuable assets rather than earning cash from a working business.
UK-based Yellow Cake, a company that stockpiles uranium, was the third name facing potential exclusion. Meanwhile, Turkey’s Lydia Holdings and Taiwan’s Center Laboratories were the others on MSCI’s watchlist.
How did MSCI choose firms to exclude?
MSCI outlined the red flags that could land companies on delist watch.
Firms whose operating assets make up more 50% of total assets advance to the next scrutiny level.
Review of five financial ratios covering operating asset intensity, expense intensity, cash flow, fair-value changes and how much a company leans on outside financing to grow.
Firms that fail to clear at least four of those five hurdles will face MSCI exclusion.
Michael Saylor’s Strategy, which holds 840,447 BTC worth roughly $53.18 billion and a free-float market value of $23.93 billion, would have tripped all five ratios on its FY2025 filings.
Ethereum treasury firm Sharplink was placed on the public watchlist because it failed the screen only once. Two consecutive annual failures translate to exclusion.
What does MSCI exclusion mean for firms?
MSCI inclusion could be worth anywhere between $2 billion and $2.8 billion in inflows for firms that make the cut. Those that get left out could also face funds flowing in the opposite direction because funds that mirror MSCI benchmarks would have to dump shares they are no longer allowed to hold.
This is MSCI’s second run at the problem in under a year. An October 2025 consultation went straight at “digital asset treasury” firms, targeting any company with at least half its assets in crypto.
MSCI dropped the crypto-specific cutoff in February after investors questioned whether a plain asset test could tell an operating company apart from an investment vehicle. The new framework swaps that single threshold for the broader ratio-based screen, which is why it sweeps in a uranium holder alongside the Bitcoin firms.
Strategy has already argued its own case. In a December 2025 letter, the company insisted it runs an active enterprise, building Bitcoin-backed credit instruments and operating analytics software, rather than sitting on a static pile of coins.
Feedback closes September 30
Nothing is settled. MSCI is collecting comments from market participants through September 30, 2026, with a decision expected around October 16.
Any changes it does adopt would take effect at the November 2026 Index Review, and MSCI has stressed the consultation guarantees none of its proposals. For now, neither Strategy nor Metaplanet has been removed.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

JPMorgan Chase has reportedly closed Polymarket’s bank account over regulatory concerns. This comes months after chief executive Jamie Dimon said the bank might one day sell prediction-market products of its own.
This contrast points to an emerging issue for Wall Street. Banks recognize that they can make money with the prediction market, where individuals invest in events like elections, sporting events, and others using real money. But entering a market like this implies taking on some legal risks. JPMorgan seems to want none of Polymarket’s involvement, even if the CEO has thought about doing so himself.
A CEO who was curious, then cautious
Dimon conveyed on the CBS Evening News to Tony Dokoupil late in March that it was “possible one day” to offer prediction market services at JPMorgan. He, however, clarified that the bank would not engage in predicting events in sports or politics and would keep “strict rules around insider information.”
When questioned on whether prediction markets represented gambling or investing, Dimon asserted that most customers saw it as “more like gambling.”
Essentially, both points of view are now found within the same financial institution. Closing the account at Polymarket while keeping the possibility of developing an in-house version of the same product demonstrates which side has won out in terms of compliance considerations.
Washington turns up the heat
Polymarket is facing increasing pressure in Washington too. On May 22, House Oversight Committee Chairman James Comer launched an investigation regarding possible insider trading on Polymarket and its competitor Kalshi. He has sent notices to Polymarket CEO Shayne Coplan and Kalshi CEO Tarek Mansour asking how each of the platforms checks account holders, what geographic limitations are imposed, and how they set alerts in case of any unusual betting.
The letters from Comer mentioned a New York Times report that revealed over 80 Polymarket accounts involved in making bets with questionable timing. Several wagers happened within hours before the US and Israeli military action against Iran was made public.
He also mentioned a federal indictment made public on April 24 that charges US Army Master Sergeant Gannon Ken Van Dyke with using classified information from the operation that seized Venezuelan President Nicolás Maduro to place bets that made over $409,000.
Comer wrote, “This growing pattern of insider trading activity on prediction market platforms indicates that Congressional action may be necessary.”
From the perspective of a lender that is unwilling to accept risks, this is exactly the type of headline that is better avoided.
Predatory-marketing allegations add to the pile
Concerns about insider trading are not Polymarket’s only regulatory challenge. As reported earlier by Cryptopolitan, a council has launched an investigation into claims that the platform has utilized predatory ways to market to young traders.
In summary, these external pressures account for JPMorgan’s preference to pull back rather than financially support Polymarket despite the industry’s promising prospects. JPMorgan is not ruling out prediction markets altogether. It is just distancing itself from a company that entails significant regulatory and reputational risks.
What to watch next
The immediate question is whether other major banks follow JPMorgan out the door and whether Polymarket can secure banking services elsewhere while the Oversight Committee investigation continues.
Comer has asked Coplan and Mansour for documents detailing how their platforms police accounts and geographic restrictions. Their responses could help determine how far Congress takes its scrutiny.
Dimon’s “possible one day” also remains significant. If JPMorgan eventually launches its own prediction-market product, it is likely to be built on the bank’s terms, with tighter controls and far less exposure to the sports and political betting that has attracted regulatory attention.
Date
Event
Jan. 3, 2022
The U.S. Commodity Futures Trading Commission (CFTC) announced a $1.4 million settlement with Polymarket over operating an unregistered facility offering event-based binary options.
Nov. 24, 2025
Polymarket US received an amended CFTC designation order, moving its U.S. operation toward a regulated designated-contract-market framework.
Oct. 2025
JPMorgan told Polymarket it needed to find another banking partner, according to Financial Times reporting. The bank cited regulatory concerns.
March 31, 2026
JPMorgan CEO Jamie Dimon said the bank could potentially offer prediction-market services, while indicating restrictions would apply to areas such as politics and sports.
April 2026
The CFTC sued the state of Wisconsin, arguing that the state’s attempts to regulate federally regulated prediction markets conflicted with federal derivatives law.
May 2026
The U.S. House Oversight Committee opened an investigation into potential insider trading on prediction-market platforms, including Polymarket and Kalshi.
June 22, 2026
The New York City Council held a hearing on legislation addressing prediction-market trading by city policymakers and access to nonpublic information.
Aug. 11, 2026
The New York City Council launched a probe into alleged predatory, deceptive, or abusive marketing practices by prediction-market platforms, including Polymarket.
Aug. 14, 2026
JPMorgan ended its banking relationship with Polymarket, while maintaining other links with the company and potentially considering an IPO role.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

TradFi trading kept up the pace in August, with a new focus on memory stocks and semiconductor companies. On-chain trading keeps shifting toward equities and metals, as pure crypto trades remain slow.
Contracts for tradfi assets like equities and metals are still attracting more active traders. Multiple platforms try to expand the representation of equities, metals, commodities, and securities, as token and pure crypto trades lose their appeal.
Crypto exchanges and decentralized platforms are now starting to reshape the way equities are traded, reaching a new audience of decentralized traders. As Cryptopolitan reported earlier, HIP-3 is already reflecting the shift, recently breaking above $4B in open interest.
Recent on-chain trading retains the overall trend from July. MEXC noted the shift to AI storage and semiconductor stocks, which remain some of the most active contracts.
Decentralized perpetual futures volumes also expanded. CoinGecko research shows volumes for TradFi assets grew 117 times over the past 18 months, becoming one of the major trends for 2026.
TradFi conviction shifts to semiconductor stocks
Semiconductor stocks are almost perfect for crypto traders. For the year to date, the sector gained 42.6%, outperforming the S&P 500 by 300%. At the same time, the semiconductor trade was extremely volatile, with periods of sharp corrections.
Semiconductor stocks outperformed the S&P500 in 2026 to date, providing both general direction and short-term fluctuations that appealed to on-chain perpetual futures traders. | Source: Yahoo Finance
After a turbulent period on the South Korean market, the country’s sector got another boost from a state-backed $3.5B fund to build a more resilient semiconductor supply chain.
As a result, semiconductors and memory stocks drew in conviction bets from HIP-3 traders. South Korea’s Hynix (TradeXYZ: SKHX) was the most actively traded asset in the sector, with $515.6M in open interest. The stock traded at $1,172, tracking the price on South Korean exchanges.
The South Korean version of SK Hynix, SKHX, was the most active equity contract on HIP-3. | Source: Hyperliquid
Whales have placed multiple high-conviction bets on both the short and long sides, leading to the past day’s 68% rally. Since HIP-3 has no limitations, SKHX and other stocks trade with no interventions and a different price discovery mechanism compared to centralized markets.
The SKHY contract, reflecting the SK Hynix American Depository Receipts, was also among the most active futures, with $224.59M in open interest. SKHY traded around $165, reflecting the US price range of the stock. The shares invited a balanced mix of high-conviction positions, split equally across short and long bets.
Sandisk (Nasdaq:SNDK), Nvidia (NVDA), and Micron Technology (Nasdaq:MU) were also among the top 15 contracts on HIP-3.
TradFi trading promises a boost to centralized exchanges
Centralized exchanges lagged behind Hyperliquid but started offering more aggressive tokenization and perpetual futures. In the past quarter, Binance led an aggressive expansion in offering equities, metals, and other TradFi assets.
Recently, Gate also marked a shift in trading activity, with new records in trading equities and precious metals. In the past week alone, TradFi trading on Gate jumped by 55%, according to recent analysis. On Gate, SK Hynix also served as a proxy for the semiconductor narrative, contributing the bulk of recent volumes, or $4.2B for the week of August 3-10.
Based on CoinGecko research, TradFi volumes reached $1.45T in the first half of 2026, with a significant boost from US equities.
The recent spikes in activity show that on-chain traders manage to react quickly and pivot in their strategies. Even with volatility, trading TradFi assets offers significant short-term gains, with improved liquidity compared to digital tokens or altcoins.
The smartest crypto minds already read our newsletter. Want in? Join them.

Apple is working on an AI model aimed at China’s market, creating a localized version of its AI model that complies with the rules and regulations of the country. The move comes reportedly with support from Alibaba as US and Chinese AI models increasingly close the capability gap.
Apple’s decision to develop a China-specific model could be a necessity in order to remain competitive in the region with specific regulations. More importantly, this will demonstrate that even successful American tech firms have to cooperate with Chinese firms and use local infrastructure to implement their AI services in the country.
Apple Intelligence clears China’s filing system
There is documentation to support the action. On July 15, the Chinese Cyberspace Administration published a list of seven newly registered generative AI services for mobile devices. As per China’s rules, these services should adhere to rules regarding the use of training data, personal information, and prohibited content.
The 2023 Interim Measures for the Management of Generative Artificial Intelligence Services require providers to “adhere to socialist core values” and comply with China’s personal information protection rules.
This makes localization more than just a case of translating from one language to another. Apple must ensure that its product is compliant with the Chinese regulatory requirements, which a local partner can assist with by providing knowledge of the technology and how to comply with the regulations. Though registration is an important regulatory hurdle that needs to be passed, it doesn’t mean that Apple gets unrestricted approval for implementing all features of Apple Intelligence.
In June, Apple had recognized the issue in a previous announcement, stating that their new Siri AI and other Apple Intelligence functionalities “will not be available in China while Apple works through regulatory requirements.” However, the July application indicates that there has been progress toward a roll-out.
A gap measured in months, not years
The other half of the story is capability. Alibaba is now among the companies in the top tier of Stanford’s Arena rankings, where models are evaluated through human voting. Stanford’s 2026 AI Index says, “The U.S.-China AI model performance gap has effectively closed.” As of March 2026, the leading US model was ahead by just 2.7%.
The US still leads in private AI investment. Stanford puts US private AI investment at $285.9 billion in 2025, more than 23 times China’s $12.4 billion. But Chinese models are now competitive across many applications.
CSIS comes to the same conclusion. In its report, it states that recent Chinese models like DeepSeek V4-Pro or Qwen3.7-Max are only “months, not years, behind U.S. frontier models.” The report mentions that, according to a U.S. government evaluation, DeepSeek V4-Pro is only 8 months behind leading U.S. models.
For Apple, this simplifies the choice. When the performance difference is measured in months rather than years, regulatory approvals, infrastructure, and adaptation to the local market become much more practical problems.
Why localizing now beats waiting
Hardware politics push in the same direction. The US approved conditional exports of Nvidia’s H200 chips to China, but Chinese authorities have restricted domestic AI companies’ access to them. A Brookings analysis said Chinese authorities “have not allowed domestic AI companies to purchase any H200 chips.”
The dispute illustrates the problem for US companies caught between Washington’s export controls and Beijing’s push for self-reliance.
CSIS says export controls have accelerated that shift, arguing their “principal effect has been to accelerate the adoption and use of indigenous equipment and products.”
For US companies competing in China, relying on local models, computing resources and partners can therefore be more practical than depending on technology whose availability may change with the next policy decision. Investment leadership and market access are becoming two different questions.
What this means for other US firms in China
Apple’s approach could become a template for other US technology companies that want to remain in China: maintain a separate AI stack for the market, work with domestic partners and build around local regulatory and infrastructure constraints.
Microsoft’s retrenchment shows the alternative. Reuters reported that the company has sharply reduced its China operations while maintaining a narrower presence serving Chinese companies with global operations. At the same time, Chinese open-weight models are becoming viable alternatives where US frontier models are unavailable or difficult to deploy.
That leaves US companies with a difficult choice. Staying in China may require greater reliance on Chinese models and infrastructure. Refusing that dependence could mean giving up customers to domestic competitors.
Apple’s China-specific model therefore matters beyond the iPhone. It offers an early glimpse of a more fragmented AI industry: one architecture for the US and much of the world, another for China, with local models and infrastructure increasingly becoming part of the cost of doing business.

If you're reading this, you’re already ahead. Stay there with our newsletter.

A Washington court has ordered Kalshi to stop offering most of its event contracts in the state, as prediction markets become one of crypto’s busiest trading arenas.
According to Pew Research data, the period between July 2024 and the first five months of 2026 saw cryptocurrency contracts contributing to approximately 20% of Polymarket’s trade volume and around 7% of Kalshi’s.
Data collected by Artemis indicate that the contribution of cryptocurrency-related contracts in prediction markets is still significant. For example, for the week ending August 9, 2026, crypto contracts amounted to about 20.3% of Kalshi’s huge $8.11 billion worth of trade, while Polymarket managed to achieve 10.2% of its volume of $1.78 billion due to transactions involving cryptocurrencies. Whenever a state intervenes in Kalshi’s offerings, it impacts the legal discussion around how far the prediction market and its connection to crypto can advance.
A gambling ruling, not a trading one
In a news release, the Washington Attorney General’s (AG’s) office said Judge John McHale of King County Superior Court established that Kalshi appeared to have violated Washington’s Gambling Act and Consumer Protection Act, as it conducted illegal gambling activities, as noted by the court. The judge provided Washington with a preliminary injunction on July 20, stating that the state was likely to win the case and rejecting Kalshi’s argument that federal commodities legislation overrode Washington’s gambling statutes.
On August 13, McHale issued an order instructing Kalshi to stop conducting any operations related to sports, elections, politics, entertainment, culture, technology, science, and “mentions,” through which the public could bet on whether public figures uttered certain words.
Kalshi has to implement an IP-address and residency geofence system by August 19, after which a multi-source system must kick in by September 2, restricting access to Washington users from various markets.
“Kalshi has gotten rich promoting wagers on sports, elections, natural disasters, events related to the Iran War, and more,” said AG Nick Brown, adding that Washington would continue enforcing the law and make sure that the company would be held “accountable for misleading consumers.” According to his office, the Kalshi ad showed a person texting that he “found a way to bet on the NFL even though we live in Washington,” which proved that Kalshi was aware of its circumvention of the law.
What Kalshi keeps, and how it got here
Some of Kalshi’s businesses remain unaffected by the ruling of the court. It has been reported by the AG that Kalshi is able to offer contracts related to commodities, climate, economics and finance in Washington. Regulators note that the categories that have been banned make up a large share of Kalshi’s income, which largely relied on sports betting.
The Washington decision followed just a few days after a court case in Utah ended unsuccessfully for the federal government. U.S. District Judge Robert J. Shelby ruled on August 4th in favor of the state of Utah and issued a summary judgment stating Utah’s laws against gambling are applicable to Kalshi’s sports contracts.
Why the crypto market is watching a state courtroom
Kalshi is a Designated Contract Market under the supervision of the Commodity Futures Trading Commission (CFTC) and claims that federal derivatives laws provide the CFTC with sole jurisdiction while state gambling laws do not apply in this case. However, state authorities hold a contrary opinion. Along with Washington, lawsuits against Kalshi have also come from Massachusetts, Michigan, Nevada, and New York, where regulations have been passed that challenge Kalshi. In turn, the CFTC has filed lawsuits against nine states to assert its federal authority over Kalshi.
On June 12, the CFTC filed a lawsuit against New Mexico, asserting that federal law gives it primary jurisdiction over event contracts and seeking to prevent New Mexico from enforcing its gambling regulations with respect to CFTC-approved contract markets. On July 14, it intervened in Michigan, stopping the change to a Kalshi regulation which would have voided past trades conducted by Michigan residents and instructing the company to honor any open trades.
The CFTC described the cancellation of already executed trades as unprecedented and threatening to undermine the integrity of the derivatives market. The conflict escalated on August 11, when the CFTC exercised emergency authority after Kalshi notified the commission of a market emergency arising from New York Attorney General Letitia James’ July 31 lawsuit. The CFTC ordered Kalshi to continue operating in accordance with the Commodity Exchange Act’s core principles. New York’s suit seeks to stop Kalshi from offering event contracts nationwide and seeks more than$36 billion in damages.
According to Token Terminal, Kalshi recorded $38.646 billion in notional trading volume in July 2026, while Galaxy Research figures cited by Cryptopolitan put prediction markets above $150 billion in cumulative lifetime volume. Macquarie Equity Research estimated in July 2026 that annual prediction-market trading volume could reach approximately $1.5 trillion by 2030.
Because these platforms increasingly rely on stablecoin settlement and crypto-native infrastructure, whether courts treat them as federally regulated derivatives or state-level gambling will help determine how far crypto’s trading rails can expand.

Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.