Don't Be Fooled by the Summer of AI Hype

A flurry of announcements and claims from major AI labs, including Anthropic and OpenAI, has driven a wave of media attention regarding incipient artificial general intelligence (AGI). Anthropic claimed its Claude Mythos model could identify software vulnerabilities better than most security experts, while both companies later disclosed hacking incidents involving their models and claimed major mathematical breakthroughs. Additionally, former Anthropic engineer Jacob Coxon went viral after alleging that the companies are rushing recklessly toward self-improving superintelligence. Despite breathless coverage framing these events as monumental technological leaps, domain experts present a far more grounded reality. Cybersecurity professionals point out that the publicized hacking incidents stem primarily from basic security negligence at OpenAI rather than autonomous models going rogue. Similarly, mathematicians initially impressed by claims surrounding OpenAI's Astra chatbot found that the results were less novel than advertised upon further examination, raising questions about marketing hype versus genuine breakthroughs.