As cited
Copy frozen at (site build).
ai security
UK AI Safety Institute Warns of Risks in Unreleased Models
The United Kingdom's artificial intelligence (AI) Safety Institute has warned that frontier artificial intelligence (AI) models present measurable risks before public release, with safety testing currently lagging behind development. The agency has renewed its call for voluntary safety commitments from AI companies to address pre-release exposure.
Why it matters: AI developers and regulators need to establish safety gates before model deployment, as testing gaps create unknown hazards for organizations adopting frontier AI systems.
- First seen by Cybersecurity Tracker