Safety & governance Policy and regulation
GAO outlines policy options for AI in cybersecurity
A new U.S. Government Accountability Office assessment examines AI-enabled attacks and defenses and presents 16 policy options across four potential goals. [1][2]
What changed
GAO reviewed research and agency material, interviewed stakeholders and convened experts. Its options address guidance, research, lifecycle governance and workforce capability. Examples include broader testing environments, output validation and human authorization for consequential actions. [1][2]
Why it matters
The report gives policymakers and security organizations a structured menu for considering AI adoption while acknowledging gaps in evaluation, data quality, privacy and organizational readiness. [1][2]
Who is affected
U.S. legislative and government bodies, standards organizations, security teams and organizations acquiring AI-enabled security tools. [1][2]
Key dates
What it means in practice
Use the assessment to examine governance and evaluation questions across a tool's lifecycle. The listed alternatives are policy choices for consideration, not newly imposed compliance obligations. [1][2]
Scope and open questions
The assessment is not a law, product certification or guarantee that AI improves security. GAO describes a non-exhaustive set of options with tradeoffs and notes that widely accepted evaluation criteria remain lacking. [1][2]
Sources and verification
Catalog evidence checked How we verify records
- Primary source AI and Cybersecurity: Technologies Can Magnify Cyberattacks and Enhance Cyber Defense
- Primary source GAO-27-108458 full report
Related profiles
Safety & governance
i.AI Liverpool youth employment support partnership
UK Incubator for AI and Liverpool City Region Combined Authority
Engineers from the government Incubator for AI will work with the Combined Authority and six local councils. The project will explore connecting information held by public services to help advisers identify young people needing support. Work is…
Safety & governance
GAO AI chatbots and mental health spotlight
U.S. Government Accountability Office
The report distinguishes general-purpose chatbots from purpose-built applications. It discusses overly agreeable responses, potentially ineffective guardrails and the limits of developer claims. GAO states that, as of November 2026, FDA…
Safety & governance
OpenAI Usage Policies October 2026 update
OpenAI
The October 21 changelog identifies the effective date. The current policy describes restrictions on preparing training material for substitute models and approval requirements for direct model-access intermediaries.