AI Safety

New York opened registration ahead of the RAISE Act’s January 1, 2027, effective date, requiring covered frontier AI developers with more than $500 million in annual revenue to register and meet safety and transparency obligations. Contractors supporting those developers may see demand for compliance documentation, incident response, and safety-assurance services.

The White House Accord on Super Intelligence encourages participating AI companies to obtain independent safety audits, but leaves auditor qualifications, access, and review standards unspecified. The voluntary commitments may create demand for AI assurance services, although no solicitation or funded contract is identified.

Reps. Don Beyer and Sara Jacobs plan to release an initial AI safety legislation draft by the end of 2026, with House floor readiness targeted for the first quarter of 2027. Potential provisions could shape future contractor expectations around AI testing, safety reporting, evaluations, liability, and export controls, but no current procurement requirement or compliance deadline has been identified.

Rep. James Walkinshaw says the next Congress should sustain U.S.-China discussions on AI risks and develop safety testing for frontier models. For federal technology contractors, the comments point to possible future scrutiny of testing, risk management, and cybersecurity, but establish no requirement, deadline, solicitation, or funding.

TÜV-Verband and TÜV AI.Lab propose mandatory independent testing of frontier AI systems before deployment and during development, including evaluator access to training processes and safety documentation. The proposal is not an enacted requirement or procurement opportunity, but could shape future demand for AI evaluation, testing infrastructure, and compliance services in Germany and Europe.

NIST’s CAISSI and Scale AI released DualChem, a 600-question benchmark designed to assess AI model performance and safety in ambiguous dual-use chemistry scenarios. The CRADA-based work is not a solicitation, contract award, or mandatory standard, but it offers contractors and agencies a concrete reference point for AI testing and evaluation needs.

The Federal Trade Commission is reportedly developing an investigation involving Anthropic and OpenAI as Washington scrutinizes AI safety-incident disclosures, lobbying, and proposals to preempt state regulation. No contract action or new requirement is identified, but government AI contractors may wish to assess reporting commitments and prepare for potential federal–state policy differences.

OpenAI is advocating for stricter national AI safety regulations, reflecting growing industry concerns about AI risks. This could result in significant procurement changes, including compliance requirements for AI vendors and contractors as states and federal agencies prioritize safety and cybersecurity assessments.