The op-ed says a shadow hangs over the Fable shutdown. A reader names who held the GPT-2 card

An essay weighs three readings of the government switching off Anthropic's Fable: genuine safety, political vendetta, or corporate competition — and refuses to pick. Then the thread supplies the fact that complicates all three.

In brief

The piece notes OpenAI's old 'too dangerous to release' GPT-2 framing as a marketing precedent. A reader points out Anthropic's Dario Amodei co-authored that GPT-2 paper — the same playbook, the same person, two companies.

Another voice, ex-Google and Meta, delivers the line that outlasts the news cycle: when you call for government power over models, it won't always be the government you wanted.

Three readings, no verdict

A shadow over the Fable shutdown — the essay's hero image
A shadow over the Fable shutdown — the essay's hero image

The essay lays out the facts of Friday, June 13: at 5:21 PM ET, with Anthropic reportedly close to an IPO, the U.S. government directed it to disable Fable 5 and Mythos 5 over a cyber-jailbreak concern, nominally for foreign nationals worldwide — a scope impossible to enforce, so access was pulled. The author then holds three explanations side by side without choosing. It could be genuine safety; he concedes GPT-2's "too dangerous to release" fears aged better than the mockery they drew. It could be a political vendetta, given a Friday-evening timing that smells of market impact. Or it could be corporate competition. His real warning is the precedent: this "may be the beginning of governments restricting the availability of strong LLMs to the public."

The thread immediately clocked the recursion — and a reader named it:

"A summary of HN comments, posted on HN to generate even more comments. Very meta." — a reader on the feedback loop

The fact that punctures the "both sides" frame

The essay treats "too dangerous to release" as a card both camps have played. A reader with a longer memory collapsed that symmetry into a single name:

"But at this time Dario was at OpenAI and was a co-author on the GPT-2 research paper… The 'too dangerous to release' approach has been him the whole time, at both companies." — a reader tracing the GPT-2 lineage to Anthropic's CEO

And a tester punctured the premise that Fable was uniquely locked down:

"the guardrails were rather sloppy and I managed to do things with Fable that really should not be possible… threat scenario analysis went out the door." — a reader who probed the guardrails firsthand

The line that outlasts the news cycle

The most durable comment wasn't about Fable at all. It was about the wish, common in the thread, that a friendlier administration had been holding the off switch:

"when you cry out for government involvement, it's not always going to be coming from the government you personally wanted. This is the whole problem with government involvement!" — a reader formerly at Physical Intelligence, Meta and Google

The precedent under the shadow

Whatever the motive — safety, politics, or a competitor's nudge — the mechanism is the part that should unsettle people. A model that companies and individuals had legally licensed was switched off by directive, with no statute, no lawsuit, and a "foreign nationals" carve-out that makes the safety rationale hard to take at face value. The essay's three theories may never be resolved. The capability they share an answer to is the real story: the off switch exists, a phone call can reach it, and this weekend was the demonstration.

Get breakdowns like this every day

The top Hacker News stories, unpacked — by email, Telegram, or RSS.

This is the last story in the June 14, 2026 issue.

Go to the issue archive