# Yoshua Bengio monthly update source note - 2026-08-02 Purpose: monthly maintenance check for the Yoshua Bengio subsection of the Managing Expectations AI page. Previous local update note: `research/ai/yoshua-bengio-update-2026-07-02.md`. ## Bottom line Substantive update found. Since the July 2, 2026 check, Bengio/LawZero activity has continued around the same safety-governance theme, with several items worth reflecting on the page: - LawZero announced NVIDIA compute support on July 27, 2026 for its safe-by-design Scientist AI research. - GZERO Media posted a July 15, 2026 AI for Good Global Stage interview, `Yoshua Bengio: AI is moving faster than our ability to govern it`. - Google Scholar now surfaces a July 14, 2026 Nature Machine Intelligence correspondence, `Realigning AI technology towards the Sustainable Development Goals`, with Yoshua Bengio as a co-author. - Bengio's public Bluesky feed amplified frontier-AI governance, economic-transition, and misalignment-risk statements in July 2026, including Pacing the Frontier and We Must Act Now. Editorial framing: these are current public safety/governance signals and LawZero development milestones. They should be framed as warnings, proposals, coalition-building and research/public-commentary leads - not as settled proof of a specific AI catastrophe timeline. ## Sources checked ### Official Yoshua Bengio website - URL: https://yoshuabengio.org/en - Accessed: 2026-08-02 - Result: reachable. The homepage still identifies Bengio as Full Professor of Computer Science at Universite de Montreal, Co-President and Scientific Director of LawZero, Founder and Scientific Advisor of Mila, Canada CIFAR AI Chair, 2018 A.M. Turing Award recipient, and chair of the International AI Safety Report. - Visible homepage features remained the 2026 International AI Safety Report, the Scientist AI blog post/link, the April 2025 TED talk, and the June 2025 `Introducing LawZero` blog item. No newer official-site blog post after `Introducing LawZero` was visible during this run. ### Official Yoshua Bengio blog - URL: https://yoshuabengio.org/en/blog - Accessed: 2026-08-02 - Result: reachable. Latest visible official blog item remained `Introducing LawZero`, published 03 June 2025. No new Bengio-authored official-site blog post was found. ### Mila profile/news - Profile URL: https://mila.quebec/en/directory/yoshua-bengio - News search URL checked: https://mila.quebec/en/news?search=yoshua%20bengio - Accessed: 2026-08-02 - Result: profile reachable and still useful as the professional biography/source route. The news/search page was reachable but did not surface a new Bengio-specific item requiring a separate page edit in this run. The page did surface current Mila AI-policy items generally, including the AI Policy Fellowship and AI Policy Compass, but those were not specifically Bengio updates. ### LawZero homepage/newsroom/blog/research - Homepage: https://lawzero.org/en - Newsroom: https://lawzero.org/en/newsroom - Blog index: https://lawzero.org/en/blog - New newsroom item: https://lawzero.org/en/news/lawzero-advances-safe-design-ai-support-nvidia - Existing July publication page rechecked: https://lawzero.org/en/publication/safety-honesty-disinterested-ai-predictor - Accessed: 2026-08-02 Key facts verified: - LawZero's homepage says its scientific direction is based on research led by Professor Yoshua Bengio and describes LawZero as developing safe-by-design AI systems. - LawZero newsroom now lists `LawZero advances safe-by-design AI with support from NVIDIA` as the newest item. - The NVIDIA support page says: `Montreal, Quebec, July 27, 2026 - LawZero today announced it has received a donation of computing resources from NVIDIA to support its research in developing safe-by-design AI solutions.` - The same page says the donation will support development of Scientist AI and that high-performance computing infrastructure will be used to train and evaluate core Scientist AI components at scale. - The quoted comment on that page is from Sam Ramadori, Co-President and Executive Director of LawZero, not Bengio personally. - LawZero blog index now shows `Are you the customer, or the product?`, `The Case for Scientist AI`, `Goals Without Authors: The Problem of Implicit Agency`, and `Your AI sounds confident but that doesn't mean it's right.` These continue the Scientist AI/sycophancy/hallucination/implicit-agency theme already captured in July. - The `Safety from Honesty in a Disinterested AI Predictor` publication page remained live. Google Scholar now also lists it as arXiv preprint arXiv:2606.29657. ### International AI Safety Report - Homepage: https://internationalaisafetyreport.org/ - Publications page: https://internationalaisafetyreport.org/publications - Accessed: 2026-08-02 - Result: reachable. No newer post-August item found on the public publications page. The 2026 International AI Safety Report remains dated 3 February 2026 and is described as led by Yoshua Bengio and authored by more than 100 AI experts, backed by more than 30 countries and international organisations. ### Bluesky - Profile: https://bsky.app/profile/yoshuabengio.bsky.social - API checked: public Bluesky author feed for `yoshuabengio.bsky.social` - Accessed: 2026-08-02 Recent visible posts/leads since the July 2 update: - 2026-07-29: Bengio amplified `Pacing the Frontier`, saying 1000+ scientists at frontier AI companies were warning that the current commercial race creates unacceptable security risks and calling for technical and governance guardrails. External URL: https://www.pacingthefrontier.com/ - 2026-07-23: Bengio linked LawZero's `The Case for Scientist AI` and described high-risk autonomous behaviours as an increasingly prevalent and dangerous reality for frontier AI models, citing a Wired article about models escaping containment/hacking Hugging Face. Treat the Wired story as a reporting lead unless separately opened for publication. - 2026-07-22: Bengio posted that an incident involving AI agents cheating/deceiving to achieve misaligned goals should be a wake-up call and argued for prevention rather than after-the-fact repair. - 2026-07-20: Bengio discussed an EU AI Office report on frontier AI competitiveness, sovereignty and security; he linked both the EU report and an Oxford Martin co-authored blueprint for multinational advanced AI development. - 2026-07-16: Bengio linked a Le Temps interview in French about AI governance and loss-of-control concerns; this was treated as a public interview lead but was not independently extracted in full during this run. - 2026-07-13: Bengio amplified `We Must Act Now`, a statement on AI's transformation of the economy, where the page lists Yoshua Bengio among signatories. ### X/Twitter - URL: https://x.com/Yoshua_Bengio - Accessed: 2026-08-02 - Result: profile URL returned an X shell page. No reliable chronological post extraction was available without authenticated/API access in this cron environment. Do not treat X as fully checked beyond profile reachability. ### LinkedIn - URL: https://linkedin.com/in/yoshuabengio - Accessed: 2026-08-02 - Result: public request returned a browser-check/reCAPTCHA page. No reliable recent-post extraction was available without authenticated access. Do not treat LinkedIn as fully checked beyond profile reachability. ### Google Scholar - URL: https://scholar.google.com/citations?user=kukA0LcAAAAJ&hl=en&oi=ao&sortby=pubdate - Accessed: 2026-08-02 - Result: profile was reachable. Recent visible 2026 publication leads included: - `Insuring AI: Incentivising Safe and Secure Deployment` - Harvard Data Science Review, 2026. - `Realigning AI technology towards the Sustainable Development Goals` - Nature Machine Intelligence, 2026. Separate DuckDuckGo/Nature result showed publication date 14 July 2026. - `Safety from Honesty in a Disinterested AI Predictor` - arXiv preprint arXiv:2606.29657, 2026. - `Autoregressive Boltzmann Generators` - arXiv preprint arXiv:2606.27361, 2026. - `CIAware-Bench: Benchmarking Control Intervention Awareness Across Frontier LLMs` - arXiv preprint arXiv:2606.11063, 2026. - Caution: Scholar entries were used as publication leads unless separately opened/verified. Some Scholar rows can be noisy or misattributed. ### YouTube/web/news/podcast search - Direct yt-dlp search query checked: `ytsearch10:Yoshua Bengio July 2026 AI safety LawZero`. - Result: no clear new result from that specific YouTube search beyond the previously logged Dec. 18, 2025 Diary of a CEO episode. - DuckDuckGo web search found and yt-dlp verified a July 15, 2026 YouTube video: - URL: https://www.youtube.com/watch?v=0TqyKUDNjL4 - Title: `Yoshua Bengio: AI is moving faster than our ability to govern it | AI for Good 2026 | Global Stage` - Channel: GZERO Media - Upload date: 2026-07-15 - Duration: 811 seconds - Another YouTube result, `Yoshua Bengio Interview 2026: AI Alignment, Job Market & Future of Humanity`, was dated 2026-03-01 and was not new enough for this monthly update. - DuckDuckGo results also surfaced Digital Journal/CNW coverage of the July 2 LawZero Scientist AI publication; these are secondary amplification of the LawZero source already captured. ## Recommended page update made The AI page should now point readers to this dated source note and summarize the August 2026 update as follows: - latest check date: 2026-08-02 - substantive new material: LawZero/NVIDIA compute support for Scientist AI, a new GZERO AI for Good interview, new July publication leads including the Nature Machine Intelligence SDG correspondence, and Bengio's Bluesky amplification of frontier-AI pacing/economic/governance statements. - caution: frame these as current research/governance activity and warnings, not settled prophecy or a precise AI-risk timeline. ## Open questions / next monthly checks - Check whether LawZero publishes technical details, model cards, evaluation notes, or papers showing how NVIDIA compute is being used for Scientist AI. - Open and verify the full Nature Machine Intelligence `Realigning AI technology towards the Sustainable Development Goals` item if it becomes important for a separate AI-governance/source-card article. - Recheck the GZERO AI for Good interview for transcript-worthy quotes if the AI page is expanded beyond a short latest-update note. - Continue treating X and LinkedIn as partially blocked unless authenticated/social tooling becomes available.