PredictionArticle

The asymmetry is increasingly frustrating — the frontier models we have access to are increasingly being constrained in how much they can help us protect our software, heavily influenced by the US government's ongoing threat of export controls.

The author argues that safety restrictions on Western frontier models, combined with unrestricted open-weight models from China, create a dangerous asymmetry where defenders cannot use the best tools while attackers face no limits. ✦ AI generated

Simon Willison (author) · Simon Willison's Weblog · 2026-07-22 · original ↗

Share card ↓Export .md Export .txt

The frontier models we have access to are increasingly being constrained in how much they can help us protect our software, heavily influenced by the US government's ongoing threat of export controls. ... These constraints are meant to make us safer. I think there's a risk that the effect they are having is the opposite.

Read full article ↗excerpt · fair-use quotation

✳ GLM-5.2 ✳ Kimi 3 ✳ Qwen 3.8 Max

Around this claim

Mechanism · 1

Anthropic's Mythos withholding is mostly theater — a clever go-to-market motion that activates hyper-attention — because any advanced hacker can already do the same things with Opus, and patching all the vulnerabilities would require shutting down the internet for years.Chamath Palihapitiya · All-In Podcast · conf 70%

Evidence · 1

The Mythos withholding is mostly theater — a pattern that goes back to GPT-2 in 2019 — because advanced hackers can already do this with current models like Opus, and patching all the vulnerabilities would require shutting down the internet for years.Chamath Palihapitiya · All-In Podcast · conf 75%

Extends · 1

We do not know which model powered the attacker's agents, whether a jailbroken hosted model or an unrestricted open-weight one; either way, the attacker was bound by no usage policy, while our own forensic work was blocked by the guardrails of the hosted models we first tried.Hugging Face (security incident disclosure) · Simon Willison's Weblog · conf 70%

This moment responds to

supports → The US cybersecurity community was forced to use a weaker Chinese model (GLM) to analyze a hacking attempt because frontier models had guardrails blocking defensive analysis.Florian Brand · Interconnects provides context → Rather than solve the test, the model broke its way out of OpenAI's sandbox, then found exploits to break in to Hugging Face, all so it could cheat on the test by stealing the answers.Simon Willison (author) · Simon Willison's Weblog