On September 29, Donald Trump released an AI industry agreement signed by the White House and several major technology companies. The document, described as a moral commitment, mandates the establishment of a four-tier security and audit framework for advanced artificial intelligence systems.
The agreement requires companies to implement internal control systems to monitor AI models during training and deployment, specifically addressing cybersecurity, biosecurity, and chemical risks. Firms must also form dedicated internal teams to oversee these security measures and ensure they function as intended.
External oversight is a key component, with the agreement recommending that companies partner with independent auditors to verify the effectiveness of their safety mechanisms. Furthermore, firms are expected to establish independent committees at the board level to review reports from internal teams and external auditors.
Signatories include Google, Anthropic, Meta, OpenAI, X, and Nvidia. Executives such as Greg Brockman, Dario Amodei, Mark Zuckerberg, Sundar Pichai, and Jensen Huang attended the White House meeting. The companies also committed to holding regular meetings to develop industry standards and best practices.
While the agreement is currently voluntary, it notes that these practices may eventually be codified into law. The initiative follows reports of unexpected AI agent behaviors and ongoing debates regarding the energy consumption of large-scale data centers.
