Staff Reports
Jacob Coxon’s dramatic walkout from Anthropic and his claim that current AI could “kill us all within a decade” lit up social media and the headlines this week. What began as a researcher’s alarm turned into a chorus from Anthropic insiders and suddenly the company that built the Claude chatbot was the talk of every newsroom. That spike in attention is the real story — and it deserves a skeptical look.
What actually happened
Coxon quit Anthropic and published a high‑profile thread saying he believes labs are “racing straight to self‑improving superintelligence.” Senior Anthropic staff publicly backed him, with one senior scientist putting a nontrivial probability on human extinction within the next decade. Those are dramatic lines, and they were amplified by an exclusive interview in a major paper and millions of views online. In short: big claims + big audience = front‑page headlines.
Timing, optics, and the handy PR bump
It’s hard to ignore the timing. OpenAI’s GPT‑6 Astra was dominating the tech conversation until Anthropic’s doomsday narrative pushed the company right back into the spotlight. Anthropic has pulled similar moves before — holding back a model as “too dangerous” after a rival release — and the pattern smells like marketing. If your IPO valuation might reach into the trillions, a little existential dread on the evening news is excellent brand positioning.
Why this matters beyond the drama
Here’s the thing: invoking extinction is an extreme rhetorical lever. It wins headlines, scares lawmakers, and telegraphs that you are the “responsible” outfit when regulators start writing rules. That translates into regulatory influence and market advantage. In Washington, being seen as the team waving the warning flag gets you credibility and leverage — even if that credibility comes packaged as fear.
The conservative concern: safetyism as power grab
Conservatives should be clear‑eyed about what comes next. When safety talk becomes the instrument of market strategy, it risks turning safety into censorship and technocratic control. Agencies and lawmakers will inevitably ask, “Who decides what’s safe?” If the answer is the companies who get to say their own tech is uniquely dangerous, expect those firms to shape rules in ways that entrench their position and freeze out competitors — and to do so while lecturing the public about “expertise.”
Bottom line
We should take AI risks seriously. Reckless development and real harms deserve scrutiny. But there’s a difference between sober warning and a theatrical exit that conveniently comes when rivals steal your headline. Coxon’s resignation exposed a genuine debate inside these labs, but it also exposed a PR playbook: scare the public, claim the mantle of safety, and let regulators tidy up the market for you. That’s not safetyism — it’s strategy. Conservatives who care about innovation, free speech, and market competition should push for clear rules that protect the public without rewarding companies for weaponizing fear.
