Does Anthropic Really Want AI to Stop? The Case That 'Safety' Is a Strategy in Disguise

Anthropic calls for the 'responsible scaling' of frontier AI to prevent human extinction, at least, that is what the company says. But this video offers a more cynical, strategic reading: a global slowdown would not just slow AI progress, it would freeze the market itself, lock today's front-runners into an advantageous position, and make it harder for open-source rivals to catch up to Anthropic's level.

And the timing is what draws scrutiny. In May 2026, Anthropic closed a record private raise at a $1 trillion valuation and filed to go public. Just three days later, it published a report warning that AI could spin out of control and urging governments to consider slowing the entire industry. To some it looks like caution; to others, like 'strategy disguised as safety', and the video lays out both competing interpretations.

As of May 2026, more than 80% of the code shipped into Anthropic's own systems was written by Claude, the company's own AI. Before Claude Code launched in February 2025, that number was only in the low single digits. In under 15 months, Claude went from barely writing any of the company's code to writing the majority of it.

At the same time, engineers are shipping more: reportedly eight times as much code per day as in 2024. Co-founder Jack Clark has openly said the share of AI-written code could reach 100% by 2028. If AI begins to design and build its own successors, the industry will need a rulebook for when to push ahead and when to stop, and, the argument goes, Anthropic already has one.

Anthropic's Responsible Scaling Policy triggers additional safeguards once capabilities reach dangerous levels (Capability Thresholds), with autonomous AI research and development in its sights. A similar idea nearly became law in California, but the venture firm a16z warned that the regulatory burden would fall hardest not on the industry's giants but on startups and small developers. A fine of 10 to 30% of training costs is pocket change to a trillion-dollar company but equivalent to an entire startup. Governor Newsom vetoed the bill in September 2024, but versions of it keep resurfacing across jurisdictions.

The impact also goes beyond Anthropic's own code. The company concedes that its engineers now mostly 'review' code rather than write it, and if that becomes the industry norm, human labor erodes. Data from Stanford's Digital Economy Lab shows that jobs for coders aged 22 to 25 are down about 20% from their late-2022 peak, while those over 30 have grown. Entry-level roles are being cut, and layoffs are accelerating across the tech industry.