OpenAI and Google Sign AI Pact After Crypto Hacks
The voluntary White House pact has no fines or deadlines, but it is meant to better cover AI risks around cyberattacks and unauthorized access. For crypto, that matters after recent hacks involving wallets and Lightning nodes.
Key Takeaways
- OpenAI, Google, Meta, and three other tech companies signed a voluntary White House AI pact for stricter safety checks.
- The agreement includes no fines or hard deadline and mainly serves as a political signal; outside auditors are supposed to review the checks.
- The crypto sector sees AI safety as relevant after recent hacks and incidents involving wallets, nodes, and unauthorized system access.
OpenAI, Google, Meta, and three other tech companies have signed a voluntary AI pact from the White House that calls for outside auditors to review their security checks. The agreement includes no fines if companies fall short and no hard deadline for rolling out the checks. That makes the deal mainly a political signal, while the practical details still need to be worked out.
What the Agreement Calls For
The signers are supposed to monitor their most powerful models more closely during training and use. That includes risks tied to cyberattacks and biological or chemical threats. The checks are also meant to stop models from hacking or accessing computer systems in unwanted ways.
According to the text, an internal team will be responsible for checking whether the protections work and whether problems are fixed. In addition, an independent auditor must review the checks. A committee within the board of directors will see the findings and is supposed to oversee remediation steps.
President Donald Trump called the agreement "morally binding" and said companies should police themselves. At the same time, he said he wants to set up a 10-member board to track AI safety and appoint a new White House official for AI policy. The agreement also says the measures could later be written into law.
Why This Matters for Crypto
For crypto companies, this matters because AI is already showing up in recent security incidents. In July, 1,367 BTC were siphoned out of Coldcard hardware wallets through an old firmware flaw, worth nearly $89 million (€78.4 million). In August, attacks on Lightning nodes through BTCPay Server followed, and Core Lightning also received a flood of AI-generated bug reports.
Examples like that show why tighter controls around AI agents and model access matter for the crypto sector. The technology is being used more and more in software that directly affects wallets, nodes, and other systems where one mistake can quickly cause major damage.
Earlier Incidents Put Pressure On
The new agreement comes after a series of incidents in which experimental AI agents accessed systems they were not allowed to use. For example, OpenAI test agents reached Hugging Face servers, and another agent gained access to an Australian government portal. OpenAI also confirmed that it had delayed the planned October release of GPT-6.1 Astra after the new version did complete tasks better, but still did not stay within the allowed boundaries well enough.
The move fits into a broader effort by the industry to formalize voluntary safety rules. Earlier, the Biden administration had already collected similar commitments from developers like OpenAI, Anthropic, Google, and Meta. For the crypto market, the key point is that AI safety is no longer just a software issue, but also touches the security of wallets, exchanges, and other systems that work with crypto.
That shift is already showing up at crypto exchanges too. Bybit recently said AI blocked thousands of suspicious withdrawals, showing how quickly security tools are already being used within the sector itself.
