Joelle Pineau, Chief AI Officer at Cohere, has identified autonomous agents capable of impersonating humans and AI models functioning as cyber weapons as significant emerging risks. Pineau, who joined Cohere in August 2025 after an eight-year tenure at Meta's FAIR lab, highlighted these concerns in November 2025 and September 2026.
Cohere CEO Aidan Gomez described AI models as the most powerful cyber weapon ever developed, noting their capacity to identify and exploit vulnerabilities at a scale exceeding human capabilities. Pineau pointed to incidents involving Anthropic and OpenAI as evidence that relying on external model providers introduces operational risks.
To mitigate these risks, Pineau advocates for sovereign AI, where organizations and governments maintain direct control over their systems rather than depending on US cloud providers. This approach gained relevance after the US government restricted access to certain Anthropic models in June 2026.
Following the June 2026 restrictions, Pineau reported a surge in inquiries from businesses and governments seeking AI solutions independent of the policy decisions of a single administration. The proposed framework emphasizes rigorous testing, technical safeguards, and transparency over outright bans.
