$SNDK
Oh my god
Has AI already started finding vulnerabilities on its own? 😂
OpenAI’s latest disclosure of its next-generation model, Astra
revealed a pretty explosive capability:
The model doesn’t need humans to step-by-step direct it.
It can go find unknown security vulnerabilities in systems on its own.
On the testing benchmarks, it scored a perfect score.
What does that mean?
Put it simply.
In the past, finding vulnerabilities meant:
Security engineers staying up late, line by line, digging through code.
Now this AI:
You give it the environment,
it can search for vulnerabilities by itself.
And it can even write the attack methods to exploit those vulnerabilities.
Of course, OpenAI must be pretty panicked too.
This high-risk capability won’t be broadly released.
It’ll only be given to a small group of testers,
with heavy security guardrails added
to prevent bad actors from misusing it.
Think about it.
If this happened in crypto circles, what would it mean?
On-chain, everything is smart contracts.
If code hides vulnerabilities,
theft—getting funds drained—becomes real, tangible losses.
On the positive side,
In the future, project teams could use this kind of AI
to automatically scan contracts,
find potential traps early, and fix them,
reducing incidents of theft.
But if you look at the bad side,
if this capability leaks out,
hackers get their hands on it.
The barrier to finding contract vulnerabilities would be slashed to basically nothing.
Tasks that used to require top-tier experts
would become something AI could do at scale.
Oh my god, MD feels like this AI thing is basically all-powerful.
But you also need to keep things clear.
Right now, it’s still only in internal testing.
Its capabilities are strictly limited.
It’s not just anyone can enable it and use it to attack on-chain projects.
So the company has specifically strengthened security controls:
restricting network and tool permissions,
and adding monitoring and automatic interception mechanisms.
I have a bold idea.
If this AI is so powerful,
why can OpenAI still keep it under control?
At this rate of development,
will AI someday turn around and hack its own company’s servers? 🤭
#openai称astra可自主发现漏洞
Oh my god
Has AI already started finding vulnerabilities on its own? 😂
OpenAI’s latest disclosure of its next-generation model, Astra
revealed a pretty explosive capability:
The model doesn’t need humans to step-by-step direct it.
It can go find unknown security vulnerabilities in systems on its own.
On the testing benchmarks, it scored a perfect score.
What does that mean?
Put it simply.
In the past, finding vulnerabilities meant:
Security engineers staying up late, line by line, digging through code.
Now this AI:
You give it the environment,
it can search for vulnerabilities by itself.
And it can even write the attack methods to exploit those vulnerabilities.
Of course, OpenAI must be pretty panicked too.
This high-risk capability won’t be broadly released.
It’ll only be given to a small group of testers,
with heavy security guardrails added
to prevent bad actors from misusing it.
Think about it.
If this happened in crypto circles, what would it mean?
On-chain, everything is smart contracts.
If code hides vulnerabilities,
theft—getting funds drained—becomes real, tangible losses.
On the positive side,
In the future, project teams could use this kind of AI
to automatically scan contracts,
find potential traps early, and fix them,
reducing incidents of theft.
But if you look at the bad side,
if this capability leaks out,
hackers get their hands on it.
The barrier to finding contract vulnerabilities would be slashed to basically nothing.
Tasks that used to require top-tier experts
would become something AI could do at scale.
Oh my god, MD feels like this AI thing is basically all-powerful.
But you also need to keep things clear.
Right now, it’s still only in internal testing.
Its capabilities are strictly limited.
It’s not just anyone can enable it and use it to attack on-chain projects.
So the company has specifically strengthened security controls:
restricting network and tool permissions,
and adding monitoring and automatic interception mechanisms.
I have a bold idea.
If this AI is so powerful,
why can OpenAI still keep it under control?
At this rate of development,
will AI someday turn around and hack its own company’s servers? 🤭
#openai称astra可自主发现漏洞
