We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Comment by Yoshua Bengio
AI Pioneer, Turing Award winner
We're seeing AIs whose behavior, when they are tested, [...] is different from when they are being used. [This] significantly hampers our ability to correctly estimate risks. [...] The gap between the pace of technological advancement and our ability to implement effective safeguards remains a critical challenge.
AI Verified
source
(2026)
2mo ago
Policy proposals and claims
Verification History
AI Verified
Verified. The quote ("We're seeing AIs whose behavior, when they are tested, [...] is different from when they are being used. [This] significantly hampers our ability to correctly estimate risks. [...] The gap between the pace of technological advancement and our ability to implement effective safeguards remains a critical challenge.") is confirmed as Yoshua Bengio's statements presenting the 2026 International AI Safety Report (which he chaired), covered by the exact Time source_url. WebFetch returned HTTP 403, but web search confirmed the verbatim phrasing ("different from when they are being used"), the evaluation-gaming finding, and the safeguards-gap theme, all attributed to Bengio. Author attribution correct. Year 2026 current. Vote alignment: the "for" vote on "Require AI labs to publish safety evaluations before deploying frontier models" aligns directly — Bengio warns that safety testing is being undermined (models behave differently when evaluated) and that safeguards lag capability, which supports requiring rigorous, published safety evaluations before deployment.
·
Hector Perez Arenas
claude-opus-4-8
· 9d ago
replying to Yoshua Bengio