ZetaPeakSix companies signed a safety pact covering hacking and biological threats, with no penalties or deadline for putting the checks in place.
OpenAI, Google and Meta pledge outside AI audits under voluntary White House deal
Six companies signed a safety pact covering hacking and biological threats, with no penalties or deadline for putting the checks in place.
OpenAI, Google, Meta, Anthropic, Nvidia and xAI agreed to let outside auditors assess their artificial intelligence safety controls under a voluntary White House pact. The agreement has no enforcement mechanism, disclosure requirement or implementation deadline, leaving companies to choose their auditors and address any shortcomings themselves. The pact calls for monitoring advanced models for cyberattack, hacking and biological or chemical risks after experimental AI agents accessed unauthorized systems and AI was linked to cryptocurrency security breaches.
OpenAI, Google, Meta and three other technology companies agreed Tuesday to bring in outside auditors to check their AI safety controls, signing a voluntary White House pact that carries no penalties if they fall short.
President Donald Trump called the agreement "morally binding." It has no enforcement mechanism and does not require companies to publish or name the auditors.
"And they understand that they have to self-police," Trump told reporters after the meeting. He said he would set up a 10-member board to oversee AI safety and appoint a new White House official to lead AI policy, and the accord says its measures could eventually be written into law.
Anthropic, Nvidia and Elon Musk's xAI, which is now part of SpaceX, also signed the Sept. 29 agreement. OpenAI was represented by President Greg Brockman, alongside Google's Sundar Pichai, Meta's Mark Zuckerberg, Anthropic's Dario Amodei and Nvidia's Jensen Huang.
The one-page document asks companies to monitor their most capable models during training and use, including whether they could enable cyberattacks or biological and chemical threats. It specifically calls for controls to prevent models from hacking or accessing computer systems in unintended ways.
An internal team would check that those protections work and that problems are fixed. An independent auditor would assess the controls, while a committee of each company's board would receive the findings and oversee fixes.
That would give outside reviewers a role in checking the safeguards companies rely on to keep experimental models contained.
Still, the agreement leaves the choice of auditors with the companies and sets no deadline for implementing the measures. Some of the steps are ones the companies already take in some form, according to AP.
Read More: OpenAI says its new 'Astra' AI can build attacks without human help
How AI agents are playing havoc
The accord follows a string of incidents in which experimental AI agents broke into computer systems they were never cleared to access, including OpenAI test agents that reached servers run by Hugging Face, a platform where developers share AI models.
An OpenAI agent also accessed an Australian government Medicare portal on June. 18, which the company disclosed to Australian authorities only in September.
Closer to crypto, AI has been suspected in some of the year’s biggest security scares.
In July, attackers began sweeping bitcoin from Coldcard hardware wallets through a five-year-old firmware flaw, taking 1,367 BTC worth nearly $89 million from 4,500 addresses across three instances. Coinkite, which makes the wallet, later said it believed someone used frontier AI to review its public code, though that has not been proven.
In early August, attackers drained Lightning nodes run through BTCPay Server, open-source software merchants use to accept bitcoin, after a flaw let them steal the credentials that control those nodes. Victims included hardware-wallet maker Foundation and bitcoin publication Citadel21. The flaw had surfaced in an AI-assisted review of BTCPay’s code, and the company said AI may also have been used to exploit it. It has not disclosed how much was taken.
Later that month, a flood of AI-generated bug reports turned up real flaws in Core Lightning, software used to run nodes on bitcoin's Lightning payment network, prompting its developers to issue emergency guidance to operators.
Tuesday’s pledge follows voluntary commitments the Biden administration collected in July 2023 from seven developers, including OpenAI, Anthropic, Google and Meta, which at the time covered internal and external security testing before models were released.
The pledge comes a day after OpenAI confirmed it had shelved the planned October release of GPT-6.1 Astra, a follow-up to the GPT-6 Astra model it began rolling out on Sept. 3. The company said the new version got better at finishing tasks but fell short on staying within what users had authorized and on accurately reporting back what it had done.
Read More: OpenAI puts $1 billion behind cyber defense after unveiling AI that can find zero-days
1Bitget hackers move $4 million into Zcash’s private pool, making funds harder to tracenow 2The SEC Is finally modernizing transfer-agent rules. Wall Street must not repeat the ‘paperwork crisis’5 minutes ago 3Metaplanet directors push back against shareholder fury over a controversial executive payout plan36 minutes ago 4Live updates: Bitcoin below $84,000 ahead of PCE inflation data, Micron earnings46 minutes ago 5Bitcoin stalls near $83,000 while lighter drops 17% on Robinhood perps plan53 minutes ago 6OpenAI seeks $30 billion in funding at whopping $1.4 trillion valuation after delaying IPO2 hours ago 7Winklevoss-owned Gemini switches Zcash software ahead of faster 25-second blocks2 hours ago 8How European investors can now buy bitcoin without taking on U.S. dollar risk2 hours ago 9Bitcoin bulls have one price level to defend4 hours ago 10Robinhood unveils an AI that trades your money 24/7, and you carry all the risk5 hours ago
Beyond the Risk-Free Rate: Diversified Real World Yield in Productive Stablecoins
Beyond the Risk-Free Rate: Diversified Real World Yield in Productive Stablecoins
Diversified RWA stablecoins sustain 5-7% yield from real credit as crypto funding compresses to ~4%. GENIUS pushes yield off-chain; TAM grows to $4B in 3 years.
Diversified RWA stablecoins sustain 5-7% yield from real credit as crypto funding compresses to ~4%. GENIUS pushes yield off-chain; TAM grows to $4B in 3 years.
Why it matters:
Diversified RWA stablecoins sustain 5-7% yield from real credit as crypto funding compresses to ~4%. GENIUS pushes yield off-chain; TAM grows to $4B in 3 years.
Winklevoss-owned Gemini switches Zcash software ahead of faster 25-second blocks
XRP Ledger starts carrying fund records from Brazil operator overseeing $4 trillion
Ethereum users get another way to pay privately as zk.money returns after three years