Canadian Godfather of AI Yoshua Bengio says that the biggest problem with Anthropic not releasing Mythos to public is that the company gets to …

Key Points:

• Canadian AI pioneer Yoshua Bengio has criticized Anthropic’s selective release of its AI model Claude Mythos, warning that concentrating control in one private company risks excluding many countries and companies from essential cybersecurity protections.
• Mythos, capable of detecting thousands of unknown zero-day vulnerabilities, is currently shared mainly with US-based firms and government agencies due to concerns over its potential misuse for cyberattacks.
• Bengio advocates for an international regulatory authority to oversee advanced AI technologies, emphasizing the need for global cooperation to prevent misuse and ensure fair access, especially amid geopolitical tensions and the US-China AI race.
• The US government plans to deploy Mythos across several federal departments despite ongoing legal disputes with Anthropic, while other countries and international institutions like the IMF and World Bank seek access to assess vulnerabilities in their own systems.
• Bengio also cautions about risks from open-source AI models, noting their capability to search for software vulnerabilities, and stresses the importance of including China in any global AI governance framework to address shared cybersecurity threats.