*Sep 29 White House lunch: Trump hosted ∼20 CEOs Musk (xAI), Zuckerberg (Meta), Huang (Nvidia), Pichai (Google), Amodei (Anthropic), Brockman (OpenAI), Bezos (Amazon), Nadella (Microsoft), Karp (Palantir) signed White House Accord on Super Intelligence (300 words, 1-page) commitments: own safe development, robust internal controls, internal verification team, external auditor, independent board committee, prevent hacking/accessing unintended systems* $AIN $COLLECT $BEAMX
*Trump framing: "morally binding" not legally binding "almost like a constitution"
"self-policing and group policing" "We automatically have regulation with DOJ, FBI, all of that" will name overseer/committee ∼10 people to monitor*
- *Sep 30 follow-up: EO telling all federal agencies stop saying "AI" / "artificial intelligence" use "Super Intelligence / SI" only, no longer acknowledge old terms Kratsios 60 days to draft law replacing term*
*Blumenthal (D-CT) Oct 2 at Senate Homeland Security rogue AI agents hearing: dismissed as "worse than ineffectual" same line as Hawley (R-MO) pushing product liability: "If an agent built without precautions knocks ER offline or locks customers out of bank, it's makers responsible"*
Why "worse": critics note 5 of 6 firms already made near-identical voluntary pledge Jul 2023 under Biden (testing, security, watermarking, promise bipartisan legislation) no law passed this one also leaves key terms "alignment," "robust internal controls" undefined, no enforcement, no independent measurement Gary Marcus: amounts to "We agree not to be regulated, no public voice, trust us"*
*BREAKING 🚨 Sen Blumenthal slams Trump-AI CEOs accord as "worse than ineffectual" 🛑 Sep 29 WH Accord on Super Intelligence signed by Musk, Zuck, Huang, Pichai, Amodei, Brockman marketed as voluntary safeguard Trump calls "morally binding" + self-policing via DOJ/FBI, no new rules Blumenthal: no enforceable measures, optics only echo of 2023 pledge that produced no law fear loophole stalls real regulation as agents go rogue ⚡*
*Trump framing: "morally binding" not legally binding "almost like a constitution"
"self-policing and group policing" "We automatically have regulation with DOJ, FBI, all of that" will name overseer/committee ∼10 people to monitor*
- *Sep 30 follow-up: EO telling all federal agencies stop saying "AI" / "artificial intelligence" use "Super Intelligence / SI" only, no longer acknowledge old terms Kratsios 60 days to draft law replacing term*
*Blumenthal (D-CT) Oct 2 at Senate Homeland Security rogue AI agents hearing: dismissed as "worse than ineffectual" same line as Hawley (R-MO) pushing product liability: "If an agent built without precautions knocks ER offline or locks customers out of bank, it's makers responsible"*
Why "worse": critics note 5 of 6 firms already made near-identical voluntary pledge Jul 2023 under Biden (testing, security, watermarking, promise bipartisan legislation) no law passed this one also leaves key terms "alignment," "robust internal controls" undefined, no enforcement, no independent measurement Gary Marcus: amounts to "We agree not to be regulated, no public voice, trust us"*
*BREAKING 🚨 Sen Blumenthal slams Trump-AI CEOs accord as "worse than ineffectual" 🛑 Sep 29 WH Accord on Super Intelligence signed by Musk, Zuck, Huang, Pichai, Amodei, Brockman marketed as voluntary safeguard Trump calls "morally binding" + self-policing via DOJ/FBI, no new rules Blumenthal: no enforceable measures, optics only echo of 2023 pledge that produced no law fear loophole stalls real regulation as agents go rogue ⚡*
