Yann LeCun Calls 'Claude Mythos' Zero-Day Panic 'BS from Self-Delusion'

A significant controversy erupted in April 2026 after Anthropic's next-generation AI model, Claude Mythos, was leaked from a public data cache before its official launch. Anthropic confirmed the model's existence, calling it a 'radical shift' in performance and their 'most capable to date.' The company then made a startling announcement: it would not release Claude Mythos to the public. The stated reason was that the 'Claude Mythos Preview' had demonstrated an ability to detect thousands of high-severity zero-day vulnerabilities across all major operating systems and web browsers, posing a potential security threat if widely available. Instead, Anthropic launched a restricted $100 million defensive cybersecurity program called 'Project Glasswing,' sharing the model only with a select group of partners including Apple, Google, and Microsoft. The claims were serious enough to reportedly prompt emergency meetings among top U.S. financial regulators and bank CEOs.

This decision and the alarming capabilities claim triggered a wave of discussion and skepticism on social media. Yann LeCun, a Turing Award winner and one of the 'godfathers of AI,' sharply criticized the narrative. On X, he dismissed the entire episode, writing, 'Mythos drama = BS from self-delusion.' He reposted analysis from AI security firm Aisle, which argued that smaller, cheaper models could largely replicate the vulnerability analysis Anthropic showcased. LeCun's skepticism was echoed by other prominent figures. AI researcher Gary Marcus called the threat 'overblown,' and former White House AI lead David Sacks noted Anthropic's 'habit of resorting to alarmist tactics.' Analysts suggested that while Anthropic might have a slight lead, it wasn't an insurmountable advantage, and competitors like OpenAI likely had similar models under similar restrictions.