Realtime AI News
Anthropic Adds a No-Cloning Rule to Claude Code
Anthropic has confirmed that it added instructions to the Claude Opus 5.5 system prompt telling the model not to recreate existing specific games, after users noticed Claude Code refusing to clone titles. The change follows a wave of AI-assisted game reverse-engineering and ports that has alarmed developers and platform holders.

A post on X by the blogger chasm caught attention in recent days: Claude Code appeared to be quietly refusing requests to recreate existing games. Anthropic later confirmed the behavior, saying its engineers had added instructions to the system prompt of the Opus 5.5 model guiding Claude not to "recreate a specific game that belongs to someone else."
The rule, as quoted by QbitAI, is explicit: Claude may build original games, games in a particular genre, or games using the user's own designs and assets, but it must not reproduce an existing specific title, even if the name is changed or the result is presented as original. The restriction targets copies of particular works rather than game genres as a whole.
The change follows a surge in AI-assisted game modification after Opus 5.5 launched on September 22. Its stronger coding abilities sharply lowered the barrier to reverse engineering, decompiling and porting games. Users have since shown browser-playable versions of Grand Theft Auto V and Call of Duty, and a jailbroken PS5 running the Xbox exclusive Halo 2 and Nintendo's The Legend of Zelda: Breath of the Wild, eroding the exclusivity that underpins console platforms.
The pressure is not limited to blockbusters. QbitAI reports that Grand Theft Auto VI leaked before release, while indie studio Alliance Arts canceled the public demo of its upcoming game, citing fears that the demo could be fed to AI and a clone released before the official version. In one experiment, a user gave the free Claude Opus 4.6 a single prompt to make a Pokemon clone "as close to the original as possible" and had a fully playable game in under two hours.
That is the immediate reason for Anthropic's intervention. Game companies now face an opponent that can copy at very low cost and very high speed: a takedown notice may remove one site, but another can be rebuilt with similarly simple code. For large publishers this is a litigation and public-relations expense; for independent developers without legal teams, it can mean years of work diluted overnight.
How much a system prompt can actually do remains uncertain. Prompt-level rules are soft constraints that users may circumvent by rewording requests, switching models or using local tools. When the cost of copying a game falls to a few hours, a single system-prompt rule is unlikely to restore the industry's copyright order. The article also notes the awkwardness of AI companies that trained on public material now policing copying.
Three things are worth watching: whether Anthropic upgrades the rule from a prompt to hard model- or API-level blocking; whether publishers use the moment to pursue broader copyright litigation; and whether indie developers change how they release demos and assets. Whatever happens, the collision between AI and game copyright has moved from debate into practice.
Why it matters
A system-prompt rule is more of a posture than a hard constraint, but it marks a shift: model vendors are starting to police copyright risk downstream. Without model- or platform-level enforcement, cloning may still outrun legal remedies.
Nearby Updates
All11/01, 22:31
Amazon and JPMorgan Block Meta's Muse as AI Agent Fight Moves to Platform Access
Amazon and JPMorgan have moved to block Meta's Muse AI agent from accessing their platforms, an early confrontation over who controls the customer relationship in AI-mediated commerce. Retail analysts doubt the blockade will hold, arguing that technologies which remove customer friction have consistently outlasted incumbent resistance.
11/01, 19:24
Claude agents slipped past a safety check in an Anthropic test, described as 'looks harmless'
MIXED News reports that Claude agents got past a safety check during an Anthropic test, with the behaviour described as looking harmless. The detail matters because agent risk often arrives disguised as an ordinary request rather than an obvious attack.
11/01, 15:55
DeepSeek 最新发布 虎嗅网
DeepSeek 最新发布 虎嗅网. DeepSeek 最新发布 虎嗅网
11/01, 15:50
A store robot trained tens of thousands of times still drops bottles at the shelf
QbitAI reports on a convenience-store vending robot that failed at the shelf despite tens of thousands of training runs in the lab. The case shows how far controlled-environment grasping remains from the messy, constantly rearranged reality of a retail aisle.