History says export controls never stopped crypto. The thread says this time the target is different

The argument the article makes

TechCrunch's history, From PGP to Mythos, lines up the failures: Phil Zimmermann prosecuted under arms-export rules in the early 1990s for PGP — a case he undercut by publishing the source code as a printed book — and spyware controls under the Wassenaar Arrangement routinely sidestepped by non-signatories. It then asks whether the same futility awaits the controls now placed on Anthropic's "Mythos" cybersecurity model, which Anthropic was reportedly ordered to restrict within 90 minutes after concerns about access by an entity with suspected Chinese ties.

The reason the analogy may not hold

The strongest reversal questioned the entire premise that "controls always fail" carries over to hosted AI:

"Except PGP could run on freely available hardware. What makes Mythos vulnerable is the centralization and scale of compute required and its proprietary nature. History shows that export controls fail on knowledge, but are damn effective on commercial products." — a commenter drawing the local-software versus hosted-service line

A second voice pointed out that controls have a quieter, more effective target than software — people:

"export controls are actually fairly successful when their target is employees within a company. Go to an AI hardware company like Nvidia… and you'll find that specific internal projects… can only be staffed by U.S. citizens, not foreigners" — an engineer on how controls bind employees, not code

What the controls were ever really for

A commenter who had lived through crypto export compliance offered the most cynical and most plausible reading of the whole regime:

"Having gone through encryption export control I came to conclusion that preventing export of technology was never the goal. By choke-pointing commercial efforts to export cryptographic tech the NSA got a perfect view and control lever of expertise export and transfer" — a commenter with first-hand crypto export-compliance experience

And on whether the controls even help strategically, one reader pointed at the counter-evidence assembling in China:

"Trump's export controls to China seem to be having the exact opposite effect as intended… Huawei is a good case in point, about to have a 100% domestic replacement for NVIDIA chips (& CUDA stack)" — a commenter tracking the Huawei and SMIC roadmap

Knowledge leaks; capacity doesn't

The thread lands somewhere more precise than the headline. Export controls genuinely fail at containing ideas — you can print PGP, you can re-derive an algorithm. They work much better at containing the things ideas need to become products: cutting-edge fabs, oceans of compute, and the specific people cleared to touch them. Mythos isn't a book. It's a service running on hardware in datacenters, which is the one shape of "technology" that export control has always been good at gripping — even as the same policy quietly pushes a rival to build the whole stack itself.