'Drama': Meta's AI head dismisses Claude Mythos fears
What's the story
Meta's Chief AI Scientist, Yann LeCun, has dismissed the fears surrounding Anthropic's latest model, Claude Mythos Preview. He took to social media platform X to say that the reaction is exaggerated and "Mythos drama = BS from self-delusion." The remarks come after the model was said to be able to find vulnerabilities in all major operating systems and browsers.
Model capabilities
Claude Mythos discovered thousands of vulnerabilities
Anthropic has claimed that its Claude Mythos model has discovered thousands of high-severity vulnerabilities in widely used software systems. The claims were significant enough to reportedly trigger discussions among financial regulators and large institutions. However, the company hasn't made the model publicly available. Instead, it is being shared with select organizations under a controlled initiative called Project Glasswing.
Mixed responses
Some experts believe model represents incremental progress
LeCun's skepticism has been echoed by other AI community members. Researcher Gary Marcus called the situation overhyped, while policy expert David Sacks pointed out a pattern of strong claims from AI companies. Some experts believe the model represents incremental progress rather than a major leap. However, cybersecurity firms like Cisco and CrowdStrike have said that AI tools like Claude Mythos can significantly reduce vulnerability detection time.
Industry impact
Cisco, CrowdStrike call AI tools like Claude Mythos game-changer
Executives from Cisco and CrowdStrike have hailed the technology as a game-changer, emphasizing that tasks that once took months can now be done in minutes. The debate over Claude Mythos's capabilities comes as AI companies are rapidly expanding and exploring new revenue streams. Reports suggest both Anthropic and competitors like OpenAI are scaling their offerings while balancing safety concerns.