AI Safety Critic Challenges Big Tech's 'Catastrophic Risk' Blind Spot

A public warning from former OpenAI and Anthropic researcher Jacob Coxon signals a critical fissure in the AI ecosystem, asserting that leading labs understand but are actively ignoring catastrophic risks. This escalates the safety debate beyond academic discourse into a direct challenge to the commercial strategies of firms like Google, Microsoft, and the frontier labs themselves. Coming just after major enterprises committed billions to AI integration, this whistleblower dynamic introduces a new vector of reputational and financial risk, questioning the very stability of the platforms underpinning the current AI gold rush. The core tension lies in the conflicting incentives between research-led labs and their enterprise customers. Labs are locked in a performance race, prioritizing model capabilities (e.g., FLOPS, parameter count) to win market share, creating a dynamic where safety is a cost center, not a feature. This fundamentally alters the risk calculus for enterprise adopters, who are now exposed not just to model hallucinations but to systemic, unpriced tail risks. Winners are smaller, safety-focused startups and consultancies that can now sell assurance services, while losers are the large cloud providers who become liability aggregators for these unmitigated risks. The trajectory now points toward a forced bifurcation of the AI market: high-capability, high-risk frontier models for specialized use, and a burgeoning market for verifiable, "boring" AI for regulated industries. Over the next 12-18 months, expect enterprise buyers to demand independent audits and safety SLAs, shifting leverage from model creators to institutional users. The critical variable is whether a major AI-driven corporate incident occurs first, which would trigger a regulatory crackdown far harsher than the industry