u.today
Longtime Bitcoin figure CobraBitcoin has sounded the alarm over a new security challenge facing the cryptocurrency industry as frontier artificial intelligence models become dramatically better at coding, vulnerability discovery and autonomous cyber operations.
The pseudonymous Bitcoin veteran says he is watching for the next major exploit after the arrival of increasingly capable models from OpenAI and Anthropic.
He has pointed to the fact that much of Bitcoin’s critical infrastructure is open-source and can therefore be analyzed at machine speed.
His warning comes shortly after OpenAI released GPT-6 Astra and Anthropic rolled out Claude Fable 5.1, two models designed to perform substantially longer and more complicated software-engineering tasks than their predecessors.
Increasingly autonomous systems could scour Bitcoin Core, Lightning implementations, wallets, libraries and other infrastructure for ordinary software bugs that human reviewers have missed.
And there is now growing evidence that AI is becoming unusually effective at probing the latter.
Critical cyber threshold
OpenAI says Astra is the first model it has broadly deployed to reach the “Critical” cybersecurity capability level under its Preparedness Framework.
According to the company, this means the model can, with the appropriate tools and access, discover previously unknown security flaws and develop new exploitation techniques against well-protected systems without requiring a human operator to direct every individual step.
Astra achieved a perfect 100% score on ExploitBench, a benchmark measuring whether models can turn known vulnerabilities into working exploits. Its predecessor, GPT-5.6 Sol, scored 78.5%.
On the more difficult ExploitGym benchmark, Astra reached a 42.4% exploit-development success rate .
Anthropic’s newly released Claude Fable 5.1 is specifically designed for long-running agentic coding and complicated work spanning large codebases.
AI warnings
The cryptocurrency industry has already received a demonstration of what AI-assisted vulnerability research can uncover.
On May 29, security researcher Taylor Hornby discovered a critical vulnerability in Zcash’s Orchard shielded pool.
The exploit could have allowed an attacker to create an unlimited quantity of counterfeit ZEC inside Orchard without detection.
Zcash developers launched an emergency coordinated responses
Following a major security scare involving COLDCARD wallets this summer, a group of Bitcoin developers and researchers organized what became known as the Bitcoin Red Team.
The team is AI systems alongside human review to scan large portions of the Bitcoin open-source ecosystem.