Show Notes
Like previous artificial intelligence models, Anthropic’s Mythos can process vast amounts of data to generate language, create images, and assist in software development. But unlike previous models, Mythos exhibits a very high capacity for discovering cybersecurity vulnerabilities. It is an artificial expert in cyberattack and defense. With little to no government oversight, private companies continue to develop more powerful AI systems. While the sources of the largest risks are located in a handful of countries, the impact of AI will be felt across the globe.
In this episode, Anthony Aguirre, the CEO of the Future of Life Institute, explains what makes powerful AI models such as Mythos so dangerous and how they represent a threat to every country in the world.
Then, our featured host, former Irish President Mary Robinson, speaks with one of the “godfathers of AI,” Yoshua Bengio, about why AI-related risks and challenges cannot be addressed by countries individually but require a coordinated approach.
ProHuman: Governing the Global AI Revolution is a co-production of Foreign Policy and the Future of Life Institute.
Mary Robinson, and all of our featured hosts, is a member of The Elders.
Yoshua Bengio is the founder and scientific advisor of Mila.
To read the International AI Safety Report 2026, click here. For more from the Future of Life Institute (FLI), click here. For FLI’s AI treaty builder, click here.
And for the latest from Foreign Policy click here.
More from ProHuman: Governing the Global AI Revolution
Right now, we’re all living through a grand experiment. The rise of artificial intelligence promises to profoundly impact humanity. While conversations about AI are often focused on technical issues, this transformative technology will shape all aspects of society. The question facing world leaders has been asked before: How do we govern a technology powerful enough to threaten civilization itself?
The artificial intelligence boom has already created extraordinary wealth and minted a new generation of billionaires. However, there are currently no mechanisms in place for AI’s massive profits to trickle down to the rest of us, even as it wouldn’t exist without our data or human labor. Some people and places are especially vulnerable.
Competition between the United States and China is fierce, but if we continue to see the artificial intelligence revolution as a race between two superpowers, then we’re putting humanity at risk. Cooperation is key to positive outcomes. With international government coordination, we can build trustworthy AI that protects nations’ sovereignty and their people.
Most global industries, including civil aviation, maritime transport, and agriculture, adhere to international safety standards. These standards and the behind-the-scenes organizations that set them are what make it safe for us to fly in an airplane, for example, or eat an apple. When it comes to artificial intelligence, however, there is no global standard-setting organization to ensure that AI is deployed safely and used responsibly.
For years, artificial intelligence safety research has focused on accidental risks and unintended consequences. But what if AI models were demonstrating risks that weren’t accidental at all? What if the AI agents were being deliberately deceptive? This sounds like science fiction, but it’s not. To avoid being shut off, some AI models have lied, manipulated information, tried to replicate themselves, and even threatened blackmail.