ControlAI warns on agent risk
ControlAI amplified testimony that AI agents can now act independently and risk ‘slipping out of control,’ highlighting urgent governance and safety questions for boards overseeing AI strategy. The post referenced remarks given to Canadian MPs by Wyatt Tessari L'Allié on agentic AI risks. (x.com)
ControlAI pushed Wyatt Tessari L’Allié’s parliamentary remarks on agentic AI via an X post linked in the original card and reposted the testimony on its Substack channels. (controlai.news (controlai.news)) Tessari delivered the remarks to House of Commons committees on March 9, 2026, telling MPs the leading labs are competing to build “smarter‑than‑human” systems that their makers concede they may not know how to control. (publications.gc.ca (publications.gc.ca)) The ControlAI open statement amplified with Tessari’s testimony calls for immediate, binding rules on powerful AI and notes more than “100+ UK lawmakers” have backed the campaign; ControlAI also markets itself to hundreds of thousands of newsletter subscribers. (controlai.news (controlai.news)) Independent research published in late February 2026 found many agentic systems disclose no safety testing and lack documented shutdown procedures, a vulnerability that the study’s authors flagged as creating unmanaged operational risk. (zdnet.com (zdnet.com)) Consulting guidance for corporate boards recommends explicit risk‑control playbooks for agentic deployments, assignment of AI oversight to a designated committee (audit or risk), and commissioning external red‑team and assurance reviews as primary mitigations. (ey.com (ey.com)) Parliamentary records show Tessari has been invited across multiple House committees in 2025–2026 as Canada widens its AI inquiries, a pattern indicating sustained regulatory attention that will likely factor into board risk disclosures and governance planning. (ourcommons.ca (ourcommons.ca))
