Musk concedes Anthropic leads AI: a rare public reversal that says what the benchmarks already showed

Elon Musk, now running the merged SpaceXAI, told an X user that he was "clearly wrong about Anthropic" and that "they are obviously currently the leader in AI," per the Washington Examiner. He added that no company has released a model as good as Mythos/Fable, that he expects Mythos 2 soon, and that he would not cut a competitor off in a damaging way. It's a notable reversal: in September 2025 Musk had said "winning was never in the set of possible outcomes for Anthropic."

The context sharpens the concession. Musk's own Grok 4.5 launched the same week OpenAI shipped its latest model, and by his own account Anthropic is still ahead. The article cites Artificial Analysis ranking Grok 4.5 as the "top non-Anthropic model," with three Anthropic models holding the top of the list. In other words, Musk is conceding a gap that an independent tracker also reports — which is exactly the kind of alignment between rhetoric and measurement we look for.

The piece also revisits the regulatory episode we've been following: Anthropic released Mythos and Fable publicly, then restricted global access after export controls tied to reports of suspected Chinese hackers accessing Mythos without authorization; the Commerce Department later lifted the restrictions after reviewing the technology, and OpenAI went through a parallel federal review. Note the attribution — the hacking access is described as suspected and reported, not established, and it's the government's concern being cited, not a proven breach.

Our reading: two things stand out. First, on our long-running "capability vs. framing" thread, this is a case where a competitor's boast is actually deflating hype rather than inflating it — Musk is talking down his own position, and the independent numbers back him. That's more trustworthy than the usual launch-day superlatives, and it reinforces our house rule: judge models by hard, independent measurement, not marketing. Second, the leadership picture matches what we've tracked — Anthropic and OpenAI at the frontier, everyone else chasing. The froth around trillion-dollar valuations and IPOs is real, but it's a separate question from who ships the best model. Capability leadership is fluid and Mythos 2, GPT-5.6, and the next Grok will reshuffle it again; the durable lesson is to keep score with benchmarks, not egos.