WORLDTECHAnthropic confirms Claude Code guardrails against cloning games
Anthropic has confirmed, through an employee's direct statement, that it added guardrails to Claude Code to stop users from recreating someone else's game. Tom's Hardware reports that the change is tied to the company's usage policy on intellectual property, and that users say Opus 5.5 is also blocking some reverse-engineering tasks.
What Anthropic has confirmed
According to Tom's Hardware, an Anthropic employee responded to a tweet by the developer Addy Osmani, which described a rule update to Claude Code, with a direct statement confirming that the company had added the guardrails. The employee's account is quoted in the report as saying that the company added two sentences to Claude Code's prompt on Opus 5.5 to steer the model away from recreating a specific game, because using Claude to infringe intellectual property breaches the usage policy.
The same account, as reported, says that work on original games, genre games and games built from a user's own designs and assets is meant to keep working, and that Anthropic is iterating on the rules to strike the right balance.
Reverse-engineering reports and open questions
Tom's Hardware reports that the analyst Walton suggested Anthropic may have added a separate rule to Claude Code specifically for reverse engineering. The outlet also reports that complaints about Claude Code blocking reverse-engineering work are growing, including one Redditor who said their account was suspended and later reinstated after using Opus 5.5 and a reverse-engineering tool.
The report describes a user who said that similar work had run without problems until October 31, and that switching to Opus 4.8 allowed the work to continue. If that holds, users may be able to avoid the limit by using another harness or by routing specific requests through a wrapper, though the report does not confirm this.
Tom's Hardware also notes that other reports of refusals appeared in the same week, which it says may indicate that cybersecurity-related rules are being adjusted in a similar way. The outlet points to Anthropic's existing Cyber Verification Program, in which only approved entities receive access to the full cybersecurity capabilities of its frontier models. The report presents this as a possibility, not a confirmed change.
Open questions
- Will Anthropic publish the exact wording of the new rules for Claude Code?
- Are the cybersecurity-related refusals linked to the same rule changes?