Claude locks more doors: games, reverse engineering, bigger models
Follow-up to Sunday Oct 25: AI labs lock down hacking skills, and open models followRemember last Sunday's token on labs locking down hacking skills? Then, a tool called REA was blocked by Claude one day after launch. Now the walls are higher. Tom's Hardware and Mashable report that Claude now blocks attempts to recreate or clone video games.
Users say the block goes further: normal reverse engineering and even basic troubleshooting get refused. Reverse engineering means taking apart software to see how it works, often for repair, security or research. Separately, a post on X says Anthropic blocks users for anything touching the training of larger models. It got 350 points and 206 comments on Hacker News, but it is a user claim, not an official rule we have seen.
Anthropic writes these limits itself, with no public process. Expect more complaints from developers, and maybe some changes if legitimate work keeps getting blocked.
When the builders restrict their own AI this hard, it is fair to ask: what are they seeing that we don't? And when the rules are private, nobody outside can check if they are fair, too strict, or too weak.
