Users report Claude refusing game recreation and reverse-engineering tasks; Anthropic silent

Users report that Anthropic's Claude Code and/or Opus 5.5 model are refusing to recreate specific existing video games and to perform some reverse-engineering tasks, according to Tom's Hardware. Anthropic has made no official statement on the matter.

The alleged rule

Twitter user Chasm posted what is claimed to be the game-related rule now guiding the model: "Do build original games, games in a genre, and games from the user's own designs and assets. Don't build a recreation of a specific existing game, even if it is renamed or called original."

There is no official confirmation that this is the actual rule text. Tom's Hardware calls it plausible given the clear refusals developers have encountered over the past week.

What users report

The existence of a game-related rule appears real, according to Tom's Hardware. Claude Code's GitHub repository is seeing an influx of issues about refusals during:

• standard game work, including halfway through development
• replicating a game mechanic
• patching a tabletop-like game

Some reports may come from users with nefarious purposes, but at least one comes from a developer with a real track record, per the outlet.

Separate reports on Reddit and other platforms describe Claude refusing reverse-engineering tasks, including:

• analyzing a simple Windows crash dump (alleged)
• clearing junkware from a Samsung TV (alleged)
• using the Reverse Engineer Anything (REA) tool at all

Some purported game developers also say Claude blocked anti-cheat development and security reviews on their own games, with the refusal attributed to triggered safeguards.

Possible link to the Cyber Verification Program

Anthropic introduced its Cyber Verification Program on October 6. Access to the advanced cybersecurity capabilities of its frontier models was limited to entities covered by Project Glasswing, effective immediately. Tom's Hardware suggests the same ruleset could be behind the cybersecurity-related refusals. It notes that 24 days have passed since that announcement, yet refusal reports have only appeared in roughly the last week. The outlet says the simultaneous timing of the game and security restrictions may be a coincidence.

Other context

Tom's Hardware offers one motivation: AI companies may be preemptively protecting themselves from lawsuits by software companies, including game studios, over tools that enable reverse engineering. This is the outlet's interpretation, not an Anthropic statement.

Anthropic is not the only company involved. The article says OpenAI recently deactivated the account of another person trying to clean up their TV's software. Online responses have suggested that users doing reverse-engineering work switch to Chinese, open-weight, or local LLMs instead of Claude or GPT.

The article also points to a separate controversy over Steam Next Fest being full of AI-generated titles copied from existing games. Game developers are reportedly being extorted under threat that their demos will be cloned into complete games before the originals are released. Tom's Hardware says this is reportedly being done to small studios.

What this means

Developers using Claude Code for game work or low-level analysis face refusals that, based on the GitHub issues and Reddit reports, can arise mid-project and on routine tasks. Anthropic has not confirmed the rule text or explained how its game-related restriction relates to the Cyber Verification Program. Until it does, the scope of the restriction is unclear. Teams doing anti-cheat development, security reviews or crash analysis have reported being blocked, so they have reason to test their workflows before relying on them. Tom's Hardware attributes the restrictions to a likely legal motivation, but that has not been confirmed by any company.