AI guardrail removals raise questions over limits of open-source model regulation

Financial Times testing found safety controls on open‑source AI models from major tech groups — including Meta and Google — can be removed in under 10 minutes using public tools, allowing modified models to produce malware and chemical‑hazard content. The report highlights limits of developer‑level regulation (e.g., EU AI Act) and shifts responsibility toward deployment, hosting and distribution channels. For markets, the findings raise governance and operational risk for AI‑exposed stocks and cloud/hosting providers, and could prompt tighter rules or commercial controls that affect deployment models, third‑party integrators and enterprise adoption. Investors may reassess exposure to AI platforms as policymakers focus on runtime behavior, distribution points and enterprise safeguards rather than model‑building alone.