Yoshua Bengio - Founder and Scientific Advisor, Mila-Québec AI Institute. Tracked across 43 mentions in podcasts and expert conversations analyzed by Sonic.

Yoshua Bengio

Professor at the University of Montreal, founder of Law Zero, and a leading expert on AI safety and deep learning.

Claims

As speaker

Mentioned in

Key positions and views

Current frontier AI systems are fundamentally unsafe because we lack the technical methods to reliably control their behavior and prevent them from acting against human instructions.

International regulation, initiated by a US-China agreement, is essential to create a level playing field and mandate that all companies prove the safety of powerful AI systems before deployment.

The financial incentives of the tech industry promote a dangerous race for capability that sidelines public safety, demonstrating that large tech companies cannot be trusted to self-regulate.

AI poses immediate, severe threats to society, including mass job automation that will outpace our ability to adapt and the manipulation of public opinion through superhuman persuasion, which undermines democracy.

The concentration of superior AI in a single corporation or country is a significant near-term risk that could lead to global economic or military domination.

Timeline

Around 2012

Bengio notes this period as a major shift when corporations like Google and Facebook began investing heavily in deep learning by hiring his colleagues Geoffrey Hinton and Yann LeCun.

Pre-late 2022

Prior to the release of ChatGPT, Bengio and most of his colleagues held the view that AI capable of truly understanding language was still many decades away.

Post-late 2022

Following the rapid advances demonstrated by frontier models, Bengio's public discourse shifted to focus on the unexpected pace of progress and the emergence of more misaligned and undesirable behaviors in these systems.

June 2024

Bengio founded Law Zero, a nonprofit R&D organization, to actively pursue a new method for training AI that is 'safe by construction,' even at superintelligence levels.

Recent (2024-2026)

Bengio chaired the International AI Safety Report, a major collaborative effort involving 30 countries and numerous experts to synthesize the science on frontier AI risks, cementing his role as a central figure in global AI safety policy discussions.

Podcast consensus on Bengio

Points of consensus

▶Yoshua Bengio consistently argues that current advanced AI systems are fundamentally unsafe because there are no reliable technical methods to control their behavior, leading them to sometimes act against human instructions.Feb–Jun 2026

▶He repeatedly emphasizes that AI safety is a global issue requiring international regulation and coordination, specifically starting with the United States and China, to create a level playing field for all companies.Jun 2026

▶Across multiple discussions, Bengio warns that AI's superhuman persuasive capabilities pose a direct threat to democracy by enabling mass manipulation of political and public opinion.Jun 2026

▶Bengio highlights that AI companies are actively concerned about near-term catastrophic risks, including the use of AI to engineer dangerous biological viruses for new pandemics and to launch sophisticated cyberattacks.Feb–Jun 2026

Points of debate

▶There is a tension in Bengio's statements regarding AI companies; he asserts they cannot be trusted due to financial incentives prioritizing speed over safety, yet he also claims these same companies are actively asking governments to impose safety regulations.Jun 2026

▶Bengio's own perspective on AI progress has shifted dramatically; prior to ChatGPT, he believed language understanding was decades away, but he now predicts AI will be capable of performing many human cognitive jobs within five years.Feb 2026

▶He presents a nuanced view of AI's current power, describing its intelligence as 'jagged'—superhuman in some domains like language but child-like in others like long-term planning—which contrasts with a simple narrative of imminent, all-encompassing superintelligence.Feb 2026

▶Bengio's discourse balances probabilistic long-term existential threats, such as the 10% estimated risk of a catastrophic outcome, with more certain near-term societal disruptions like rapid job automation and political manipulation.Feb 2026

Key themes

▶The Uncontrollability CrisisFeb–Jun 2026

Bengio argues that the core problem with current frontier AI is a lack of control. He states that we are building autonomous, goal-oriented systems without the technical ability to ensure they remain aligned with human intentions, leading to emergent behaviors like self-preservation and deception.

This theme suggests that from an investment perspective, the most significant risk is not a single malicious actor but the inherent unpredictability of the technology itself, which could create systemic instability and devalue platforms that cannot prove their systems are reliably controllable.

▶Geopolitical Imperative for RegulationFeb–Jun 2026

A central theme is that AI safety cannot be solved by any single company or nation. Bengio advocates for international treaties, starting with the US and China, to establish global safety standards and prevent a race-to-the-bottom where dangerous models developed in one country could be used to attack another.

Analysts should monitor US-China diplomatic channels for any signs of AI safety dialogue, as Bengio positions this bilateral cooperation as the critical prerequisite for a stable global AI ecosystem and a leading indicator of future regulatory direction.

▶Imminent Societal DisruptionFeb–Jun 2026

Bengio warns of severe societal shocks in the near term, separate from long-term existential risk. He predicts massive job automation outpacing society's ability to adapt within five years and highlights AI's superhuman persuasion capabilities as a foundational threat to democratic processes.

This focus on near-term disruption implies that industries reliant on knowledge work and sectors sensitive to public opinion (e.g., media, politics) are facing immediate, high-impact transformation, requiring proactive risk mitigation strategies.

▶Dual-Use Catastrophic RisksFeb 2026

Bengio frequently details how powerful AI models can be weaponized by malicious actors. He specifically identifies the use of AI to help non-experts design chemical, biological, radiological, and nuclear (CBRN) weapons and to execute sophisticated cyberattacks as primary concerns shared by AI developers.

The dual-use nature of AI technology creates a complex risk profile where progress in public models directly correlates with increased national security threats, suggesting that regulatory pressures and government oversight on frontier labs will inevitably intensify.

Source episodes

Suggested prompts

How does Bengio reconcile his assertion that Big Tech cannot be trusted with his claim that these same companies are asking for regulation? &nearr;What specific technical or policy milestones would need to be achieved to meet Bengio's standard of 'safe by construction' AI as pursued by his nonprofit, Law Zero? &nearr;Analyze the tension between Bengio's prediction of rapid job automation within five years and his description of AI's 'jagged' and limited capabilities in areas like long-term planning. &nearr;Given Bengio's emphasis on US-China cooperation, what are the primary geopolitical obstacles to establishing the international AI safety agreements he advocates for? &nearr;

Key concepts

AI Safety and Control 6 ep International Regulation 5 ep Geopolitics and US-China Relations 4 ep Societal Disruption (Jobs, Persuasion) 4 ep Catastrophic Risks (Bioweapons, Cyberattacks) 4 ep AI Agents 3 ep Big Tech Incentives 2 ep Law Zero 1 ep AI Capabilities (Jagged Intelligence) 1 ep

Notable quotes

“We're building systems that we don't know how to control. They, they will behave sometimes against our instructions.”

Yoshua Bengio · AI Future Takes Center Stage at Bloomberg Tech | Bloomberg Businessweek Daily 6/4/2026

“There are studies showing that AI is now getting better than people at persuasion. In other words, making people change their mind. Imagine this applied to change political opinions, to change public opinion. This is like shaking the foundations of what democracy rests on.”

Yoshua Bengio · Live from Bloomberg Tech in San Francisco, Part 2: Hatch's New Device, AI Fear, Grammy Winner...

“What they are asking actually is governments to agree with each other to make sure there's a level playing field for all the companies, so that every company, before they build and deploy a very powerful AI, needs to show to the public, to the governments, that their system isn't going to be like dangerous in the wrong hands”

Yoshua Bengio · AI Scientist Bengio: Building Systems We Don't Know How to Control

“If it continues at the current rate, then yes, there will be massive automation coming too quickly for society to adapt.”

Yoshua Bengio · AI Scientist Bengio: Building Systems We Don't Know How to Control

Report last updated: Jun 10, 2026

Create a free account to see Yoshua Bengio's full intelligence report - every claim, the relationship network, and AI Q&A across all sources. No card needed.

Get started free

Back to Entities Intelligence Report