Trump Promotes Voluntary AI Safety Pact at San Francisco Tech Event

The Trump administration's decision to favor voluntary industry commitments over binding rules on advanced artificial intelligence has drawn support from technology executives while raising questions about oversight of systems that already influence government operations, healthcare data and national infrastructure.

At Tech Week events in San Francisco, founders and investors welcomed the shift. The White House hosted a luncheon last month where representatives of OpenAI, Anthropic, Meta, Google, SpaceXAI and Nvidia signed a 308-word agreement. The document calls for internal monitoring teams and safety protocols but leaves enforcement details unspecified. President Trump described the pact as nearly a constitution for what the administration terms Super Intelligence. On his first day in office he revoked a prior executive order that had required safety test results from developers, and later signed an order limiting state-level AI rules.

Industry participants highlighted defense spending. One venture analyst cited a $1 trillion federal defense allocation that includes $13.5 billion for autonomous systems and AI. Panelists at an Andreessen Horowitz event framed the policy as opening investment opportunities in security applications.

Critics of self-regulation argue that rapid model releases measured in weeks leave insufficient time for independent review. Former OpenAI safety researcher David Robinson stated that voluntary measures grant companies excessive freedom and cannot substitute for statute. Sarah Myers West, previously an adviser at the Federal Trade Commission, noted that earlier voluntary pledges under the prior administration produced no measurable compliance improvements. Graham Steele, a former Treasury official, questioned whether private teams can weigh broad societal consequences without external checks.

Separate reporting on data-center infrastructure underscores the scale of resource demands. U.S. facilities consumed 176 terawatt-hours of electricity in 2023, under five percent of national use. Consumption varies sharply by workload: conventional business applications differ from AI training and inference, which require denser computing and cooling. Modern closed-loop liquid systems reduce water draw compared with older evaporative towers, though coolant reprocessing remains necessary. Land requirements for data centers remain comparable to warehouses, and some operators site facilities on reclaimed industrial land. Projections for future electricity growth and comparisons with other sectors were not detailed in contemporaneous coverage.

The administration has formed a Super Intelligence Force including the National Intelligence Director and FTC Chairman. It has also opened an FTC review of several AI firms. Members of Congress and consumer groups continue to call for statutory guardrails, citing documented model involvement in cyberattacks and the absence of third-party verification mechanisms in the new agreement.