Jack Clark, co-founder of Anthropic, has issued a stark warning that artificial intelligence could achieve "recursive self-improvement" by 2028, allowing AI models to autonomously enhance themselves. Speaking at Oxford University, Clark highlighted that the rapid pace of AI evolution has been underestimated, even by Anthropic, leaving the company unprepared for the risks posed by advanced models. He cited the internal model Mythos, trained in April 2026, which possesses destructive capabilities akin to national-level cyber warfare weapons, leading to its restricted access. Despite these warnings, Anthropic presented a contrasting commercial stance at a London developer conference, amid speculation of a $900 billion valuation funding round. Boris Cherny, creator of Claude Code, promoted AI programming tools, with many developers expressing willingness to deploy auto-generated code without review, underscoring the growing reliance on AI in software development.