Sam Altman accuses Anthropic of using "Fear Marketing" to promote the Claude Mythos

April 23rd — Per Decrypt, OpenAI CEO Sam Altman pushed back against recent warnings about rival Anthropic’s powerful new Claude Mythos model, suggesting the company is using “fear marketing” to promote the product. Speaking on the *Core Memory* podcast hosted by tech journalist Ashlee Vance, Altman argued the “fear-based marketing” tactic aims to keep AI in the hands of “a few.” He noted: “Legitimate security concerns exist, so some arguments for this hold weight. But if the goal is ‘we need to control AI, and only we can because we’re trustworthy,’ fear marketing is the most effective pitch.” Altman added that while AI safety concerns are valid, he likened the tactic to: “We’ve built a bomb that’s about to drop on your head, and we can sell you a $100 million bomb shelter to protect your stuff — but the catch is we handpicked you as the customer.” Balancing AI’s new capabilities with OpenAI’s belief that “technology should be accessible” is “not easy,” he said. Released last month, Claude Mythos has drawn heavy attention from researchers, governments, and the cybersecurity industry for its ability to autonomously find software vulnerabilities and execute sophisticated network operations in testing. Currently, the model is only available to a select few institutions — including Amazon, Apple, and Microsoft — via the restricted Project Glasswing. Earlier this month, it uncovered hundreds of vulnerabilities in Mozilla’s Firefox browser during testing. “There will be more talk about models ‘too dangerous to release,’ but some very dangerous models must be rolled out differently,” Altman said. “I think Mythos is an excellent cybersecurity model, but we have our own plan we’re quite happy with for bringing such capabilities to the world.”