Yoshua Bengio

Des dialogues comme celui-ci me rendent optimiste quant à notre capacité d'agir ensemble pour façonner un avenir plus sécuritaire avec l’IA!

Yoshua Bengio et l’urgence d’un réveil citoyen quant à l’intelligence artificielle | OHdio | Radio-Canada Écoutez l’extrait de l’émission Pénélope : Yoshua Bengio et l’urgence d’un réveil citoyen quant à l’intelligence artificielle

Merci à Pénélope McQuade de @iciradiocanada.bsky.social de m'avoir invité à son micro ce matin pour répondre aux excellentes questions du public concernant les répercussions de l'IA sur nos sociétés.

ici.radio-canada.ca/ohdio/premie...

We are very proud to welcome Dame Jacinda to @law-zero.bsky.social's Global Advisory Council!
Her exceptional track record of leading with integrity and her experience in international approaches to technology-driven global issues will be of great value to our organization.

The Pro-Human AI Declaration The Pro-Human AI Declaration

The race to develop frontier AI is accelerating faster than safeguards can keep up, posing major risks to democracy and our societies.
I’m proud to add my voice to the growing movement of experts and organizations calling for a safer, more intentional path forward with AI.
humanstatement.org

I’m grateful for the trust of my colleagues and look forward to our collaboration to create a reliable, independent scientific assessment of AI's capabilities and impacts, providing an essential resource for citizens and policymakers around the world.

I am deeply honored to have been elected as Co-Chair of the United Nations' Independent International Scientific Panel on AI alongside Nobel Peace Prize laureate @mariaressa.bsky.social.

Anthropic CEO says he's sticking to AI "red lines" despite clash with Pentagon After the Trump administration cut it off, Anthropic CEO Dario Amodei told CBS News in an exclusive interview Friday night he wants to work with the military — but only if it addresses the firm's conc...

Current AI lacks the reliability required for high-stakes areas—companies have a responsibility to exercice their fiduciary duty to ensure their products are not used to harm humans.
www.cbsnews.com/amp/news/pen...

IA : Yoshua Bengio alerte sur "le pouvoir incontrôlé qui est en train de se développer" YouTube video by France Inter

Ravi d'avoir rejoint Florence Paracuellos et Benjamin Duhamel au micro de @franceinter.fr ce matin pour discuter des risques de l'IA et des pistes de solutions, notamment celles que nous développons @law-zero.bsky.social.
www.youtube.com/watch?v=6Zm-...

Looking forward to this panel discussing the key findings of the International AI Safety Report at the India AI Impact Summit tomorrow!

I’ll be joined by Josephine Teo, @alondra.bsky.social, Adam Beaumont and Lee Triedrich.

I am honored to have been appointed to the UN' International Independent Scientific Panel on AI.

I look forward to working alongside global experts to advance scientific understanding and ensure that international deliberations are informed by the best available evidence.
www.un.org/independent-...

At LawZero, we're rethinking the building blocks of frontier AI to create an intelligent machine that is both highly capable and safe-by-design. We’re excited to share our first blog post outlining some of the objectives and core components of our Scientist AI project. 🧵
(1/4)

I’ve found the collaborative spirit of the 100+ contributors heartening, and am grateful to have benefitted from their complementary perspectives.
Thank you to all contributors for their dedication.

(19/19)

2026 Report: Extended Summary for Policymakers The Extended Summary for Policymakers of the 2026 International AI Safety Report. The second International AI Safety Report, published in February 2026, is the next iteration of the comprehensive revi...

With all the noise around AI, I hope this Report provides policymakers, researchers, and the public with the reliable evidence they need to make more informed choices. We also have an “Extended Summary for Policymakers”:
internationalaisafetyreport.org/publication/...

(18/19)

Because no single safeguard reliably prevents misuse or malfunctions, developers are converging on "defence-in-depth,” layering multiple measures—model-level training, input/output filters, monitoring, access controls, and governance.

(17/19)

However, safeguards remain imperfect.
Attackers can still often find ways to evade them fairly easily. One initiative crowdsourced over 60,000 successful attacks against state-of-the-art models. When given 10 attempts, testers can still generate harmful responses about half the time.

(16/19)

These safeguards inform institutional risk management approaches. For example, 12 companies published or updated Frontier AI Safety Frameworks in 2025—more than double the prior year.

(15/19)

Many technical safeguards are improving. For example, models hallucinate less and it is harder to elicit dangerous responses.

(14/19)

Even areas of uncertainty carry risks that warrant attention.
For example, in 2025 multiple companies added safeguards after pre-deployment testing could not rule out the possibility that new models could assist novices seeking to develop biological weapons.

(13/19)

Wider adoption is also raising new challenges. For example, this year we discuss early evidence on how “AI companions”, which are now used by tens of millions of people, may affect people’s emotions and social life.

(12/19)

There is little evidence of overall impacts on labour markets so far, though early-career workers in some AI-exposed occupations have seen declining employment compared with late 2022.

(11/19)

Since the last Report, we’ve seen new evidence of many emerging risks. For example, AI-generated content has become highly realistic and more useful for fraud, scams, and non-consensual intimate imagery. There is growing evidence that AI systems help malicious actors carry out cyberattacks.

(10/19)

But new capabilities pose risks, notably 8 emerging risks.
Misuse:
→ AI-generated content & criminal activity
→ Influence & manipulation
→ Cyberattacks
→ Bio & chemical risks
Malfunctions:
→ Reliability issues
→ Loss of control
Systemic risks:
→ Labor market impacts
→ Risks to human autonomy

(9/19)

These capabilities are increasingly translating into real-world impact.
At least 700 million people now use leading AI systems weekly. In the US, use of AI has spread faster than that of computers and the internet.

(8/19)

But capabilities are also “jagged:” the same model may solve complex problems yet fail at some seemingly simple tasks.

(7/19)

On capabilities: AI systems continue to improve significantly.
Leading models now achieve gold-medal performance on the International Mathematical Olympiad.
AI coding agents can complete 30-minute programming tasks with 80% reliability—up from 10-minute tasks a year ago.

(6/19)

2️⃣ Some risks, from deepfakes to cyberattacks, shifted further from theoretical concerns to real-world challenges.

3️⃣ Many safety measures improved, but remain fallible. Developers increasingly implement multiple layers of safeguards to compensate.

(5/19)

This report provides policymakers with the information they need to make these decisions.

In 2025:
1️⃣ Capabilities continued advancing rapidly, especially in coding, science, and autonomous operation.

(4/19)

AI poses an “evidence dilemma” to policymakers—capabilities evolve quickly, but scientific evidence emerges far more slowly.
Acting too early risks entrenching ineffective policies, but waiting for strong evidence may leave society vulnerable to risks.

(3/19)

International AI Safety Report 2026 The second International AI Safety Report, published in February 2026, is the next iteration of the comprehensive review of latest scientific research on the capabilities and risks of general-purpose ...

Over 100 independent experts contributed to the Report, including Nobel laureates and Turing Award winners, along with an advisory panel with nominees from more than 30 countries and international organisations, including the EU, OECD and UN.
internationalaisafetyreport.org/publication/...

(2/19)

Today we’re releasing the International AI Safety Report 2026: the most comprehensive evidence-based assessment of AI capabilities, emerging risks, and safety measures to date. 🧵

(1/19)