US President Donald Trump and executives from six leading technology companies signed a voluntary artificial intelligence safety accord on 29 September, committing their businesses to internal controls, independent assessments and board oversight of advanced systems.The White House Accord on Super Intelligence covers Google, Anthropic, Meta, OpenAI, xAI and Nvidia. It establishes four layers of checks for companies training and deploying frontier models, but creates no enforcement mechanism, penalties or binding timetable for implementation.
Trump described the agreement as “morally binding” after a White House lunch with technology executives. The document, subtitled Joint Commitment on Frontier Responsibilities, leaves responsibility for operating the safeguards primarily with the companies themselves, while acknowledging that legislation or regulations could eventually incorporate them.
The signatories included Google chief executive Sundar Pichai, Anthropic chief executive Dario Amodei, Meta chief executive Mark Zuckerberg, OpenAI president Greg Brockman, Elon Musk and Nvidia chief executive Jensen Huang. Trump also signed the accord, which he subsequently published on Truth Social. ndtvprofit. com
Its first layer calls for internal controls to track model capabilities and alignment during training and deployment, with particular attention to cybersecurity, biological security and chemical threats. Companies also undertake to prevent their models from hacking or accessing technical systems in unintended ways.
A second layer requires an empowered internal team to check that controls, monitoring and detection work properly and that problems are corrected. The third calls for an independent external auditor or evaluator to assess whether those arrangements operate as intended.
The fourth assigns oversight to an independent committee of each company's board of directors. That committee would receive reports from teams operating the controls and from internal and external assessors, and ensure that identified issues are remedied.
The participating companies also agreed to meet regularly to develop standards and best practices for improving system safety. The accord does not specify how frequently those meetings should take place, require publication of audit findings or establish a common threshold for stopping an unsafe model's release. forbes. com
The document says each company bears responsibility for developing technology safely and maintaining public trust. Its provisions focus on checking whether systems behave as intended, rather than creating a government approval process for deploying new models.
The text does not name an organisation to supervise compliance across the six companies or explain how the independence of external assessors would be guaranteed. It also leaves the selection of auditors and the practical design of monitoring arrangements to participants. Reuters
Pichai compared the arrangements with established corporate financial controls, saying companies were applying familiar processes to advanced AI development. Zuckerberg described a system combining internal risk reviews, external evaluators and scrutiny by boards of directors.
Amodei offered a more qualified assessment, saying the technology carried “very real risks” and that mechanisms for addressing them remained under discussion. Huang said technological innovation and safety could advance together, arguing that companies were developing tools to improve both capabilities and protection.
Criticism centred on the reliance on corporate oversight. Alex Pascal, executive director of Harvard University's Berkman Klein Center for Internet and Society, argued that reducing harms required legal liability and regulation, alongside changes to competitive pressures driving unsafe decisions by frontier developers. AP News
Trump said he was considering a committee of about 10 people to oversee AI safety, but did not identify its prospective members or powers. That proposal was separate from the board committees described in the signed document. He also said he would appoint a new official to lead White House AI policy. Reuters
Alongside the accord, Trump signed an executive order directing federal agencies to replace artificial intelligence and AI with Super Intelligence and SI in official communications and other non-statutory documents, to the maximum extent permitted by law.
The order retains the existing statutory definition of artificial intelligence for implementation. It therefore changes federal terminology without establishing that today's systems have achieved capabilities beyond human intelligence. The directive also leaves previously issued regulations, presidential actions, contracts, grants and historical documents unchanged.
Topics
Technology