Anthropic's recent launch of the Mythos AI model has sparked significant concern among its customers, particularly regarding the model's potential to expose cybersecurity vulnerabilities. During a recent event at the Vatican, Anthropic cofounder Chris Olah revealed that the company's engineers have encountered unsettling discoveries within their AI models. Developers attending Claude Code workshops in London expressed worries about the increasing autonomy of AI tools, which raises critical questions about accountability. Cat Wu, head of product for Claude Code, attempted to reassure developers by emphasizing the system's security, attributing concerns to communication issues rather than a lack of controls. However, many developers feel sidelined as they observe AI generating code independently, further complicated by the latest models not providing clear insight into their reasoning processes. This trend raises alarms about potential skill atrophy among programmers who may become overly reliant on AI tools. While the costs of AI may eventually limit access, leading to more human oversight, the risks associated with reduced developer engagement remain evident.