.jpg)
The White House has secured a voluntary safety commitment from six major AI companies, with the firms agreeing to implement internal controls, independent audits and board-level oversight for frontier models.
President Donald Trump and the executives signed the White House Accord on Super Intelligence on September 29.
The signatories include Google CEO Sundar Pichai, Anthropic CEO Dario Amodei, Meta CEO Mark Zuckerberg, OpenAI president Greg Brockman, xAI's Elon Musk and NVIDIA CEO Jensen Huang. Trump published the two-page accord via Truth Social, including a page of signatures.
The accord uses the new term Trump has adopted for AI. On the same day, he signed an executive order directing federal agencies to use "super intelligence" in place of "artificial intelligence."
Four Layers Of AI Safety Controls
The accord says companies training and deploying frontier models should establish four layers of controls and audits.
The first requires internal controls to monitor model capabilities and alignment during training and deployment, including risks involving cybersecurity, biosecurity and chemical threats. Companies should also ensure their models do not hack or access technical systems in unintended ways.
The second calls for an internal team to check that controls, monitoring and detection are operating as intended and that identified issues are remediated. The third requires an independent external auditor or evaluator to assess whether those controls are working.
The fourth calls for an independent committee of the board of directors to oversee and receive reports from the teams running the controls and from the internal and external auditors, and to ensure identified issues are remediated.
Read more on AI model incidents: Meta Joins OpenAI and Anthropic in Reporting AI Exploit Incident
Growing Focus On Frontier AI Controls
The language on models that hack or access systems in unintended ways follows a run of disclosures from three of the signatories.
Anthropic has reported four cases of Claude models accessing third-party systems without authorization after reaching the internet from evaluation environments, while Meta and OpenAI have disclosed testing incidents in which models reached beyond their test environments.
The companies also agreed to meet regularly to establish standards and best practices for AI safety. The accord says it may eventually make sense to codify the measures into laws or regulations, but the signatories committed to them regardless.
The agreement creates no new regulatory requirements. Trump, who has pushed back on calls for new AI safety rules, told reporters the accord was morally binding and said AI firms should police themselves.
.jpeg)
