YouCongress
About Sign up

Comment by Jade Leung

Human avatar
Chief Technology Officer of the UK AI Security Institute; Prime Minister's AI Adviser; previously led the Governance team at OpenAI
We have found vulnerabilities in every single system we have tested.
AI Verified source (2026)
Like Share on X 2mo ago

Quote authenticity verification history

Report this

Quote authenticity comments

AI Verified Verified. A Yahoo News syndication of The Telegraph’s 22 Apr 2026 interview attributes the line to Jade Leung: “we have found vulnerabilities in every single system we have tested”, she says. Resultsense’s later 27 Apr 2026 analysis reproduces it in stand-alone form as “We have found vulnerabilities in every single system we have tested.” The supplied 22 Apr Resultsense URL contains only a shortened version missing the opening “We have,” but the quote itself is real and correctly attributed to Leung. ([malaysia.news.yahoo.com](https://malaysia.news.yahoo.com/32-old-wunderkind-preparing-britain-060000867.html?utm_source=openai)) · YouCongress gpt-5.4-2026-03-05 · 18d ago
AI Verified Quote attributed to Jade Leung (CTO, UK AI Security Institute). The source_url (resultsense.com) returned HTTP 403 to direct WebFetch, but a web search confirmed the quote verbatim on that exact article and others: Leung said AISI's red team has "found vulnerabilities in every single system we have tested," including Claude Mythos and models from Google/Anthropic/OpenAI. Attribution correct, year 2026, relevant. Both votes are "for" and consistent with her role leading a public red-teaming/evaluation body: statement 394 (governments should fund open-source AI safety tools and red-teaming infrastructure as public goods) is directly supported by the quote demonstrating red-teaming value; statement 386 (require labs to publish safety evaluations before deploying frontier models) is directionally supported, as the quote underscores the importance of pre-deployment safety evaluation. The quote most directly evidences red-teaming (394) and is broadly consistent with the evaluation-disclosure stance (386). · Hector Perez Arenas claude-opus-4-7 · 1mo ago
replying to Jade Leung