WSJ Article Claiming China Has Matched Anthropic Is Obvious Nonsense
Zvi's AI Roundups · Zvi Mowshowitz · 2026-06-29
Zvi Mowshowitz argues that a Wall Street Journal headline claiming China matched Anthropic's Mythos model in cybersecurity is false, because GLM-5.2 only equals Mythos on narrow, easy tasks and lacks its defining capability to autonomously discover and chain exploits at scale.
Appears in
Extraction
Topics: ai-capabilitiesus-china-ai-competitioncybersecurity-aimedia-accuracy
Claims
- The WSJ headline claiming China matched Anthropic's Mythos in cybersecurity is factually false and the article creates a heavily misleading overall impression.
- Mythos's defining capability is autonomous, large-scale vulnerability identification and exploit chaining without being directed at specific code sections — something GLM-5.2 cannot replicate.
- GLM-5.2 matches Mythos only in narrow scenarios where models are pointed at specific code subsections, a capability most capable models share.
- The AI capability gap between the US and China has not been predictably shrinking; it widened before GLM-5.2 and narrowed only on its release date.
- The WSJ and other outlets have a pattern of publishing false or misleading 'Chinese models have caught up' headlines with real policy consequences.
Key quotes
What makes Mythos special is that it can identify vulnerabilities autonomously, at scale, without being pointed at them, and can then autonomously string together a variety of seemingly unrelated vulnerabilities into full working exploits.
The overall impression of the article, even excluding the headline, is that China is steadily catching up if it has not already done so...All of this is completely wrong.
Shame upon the Wall Street Journal. I fear Gell-Mann Amnesia. If they can get something as important as this so completely wrong, what about everything else?