Seven labs sign voluntary safety commitments at the White House

The Biden administration announced that seven companies — Amazon, Anthropic, Google, Inflection, Meta, Microsoft and OpenAI — had agreed to a set of voluntary commitments on AI safety, following meetings between company executives and White House officials. Company representatives, including chief executives, appeared alongside President Biden for the announcement.

The commitments covered several areas: internal and external security testing of models before release, partly conducted by independent experts; sharing information on risk management across industry, government, academia and civil society; investment in cybersecurity to protect model weights against unauthorised access; facilitating third-party discovery and reporting of vulnerabilities; developing mechanisms — including watermarking — so users could tell when content was AI-generated; public reporting on a system’s capabilities, limitations and risks; and prioritising research into societal harms such as bias and privacy.

The document contained no enforcement mechanism, no penalties for non-compliance, and no independent verification process; it described the commitments as ones the companies “have chosen to undertake,” and the administration said it would separately pursue an executive order and legislation. That gap between rhetorical commitment and binding obligation was a repeated theme in US AI policy through the period.

The announcement was one of two industry-safety moves in July 2023, alongside the Frontier Model Forum that four of the same companies launched five days later; together they marked the point at which voluntary self-governance, rather than statute, became the default US posture toward frontier AI for the next several years, a gap Biden’s executive order that October only partially closed.