Anthropic and OpenAI are strengthening intellectual property-related restrictions. AI can still be used for game and other software development, but directly copying protected software or game content is more likely to trigger output blocks, warnings, or account actions.
The scope of AI programming tools is expanding. OpenAI data shows that Codex had over 5 million weekly users by June 2026, with applications including apps, automation projects, and game development. Anthropic disclosed that computer and math-related tasks accounted for about 35% of conversations on Claude.ai, and the proportion of GitHub projects using AI coding agents more than doubled. However, neither company has released data sufficient to accurately quantify the number of games or applications considered "AI-generated low-effort content."
Anthropic launched Claude Opus 5.5 on September 22, highlighting complex tasks and programming as its primary use cases. The company demonstrated that the model migrated a codebase containing 680,000 lines of code in less than a day. Meanwhile, Claude's content filtering system may halt certain outputs via prompt warnings, with restrictions also applied to prevent the regeneration of protected materials.
Under Anthropic's rules, users are prohibited from using its services to infringe on others' intellectual property rights. Intentionally requesting copyrighted content may result in warnings, and repeat offenders may face account restrictions or termination, though appeals are possible. In the first half of 2026, Anthropic received approximately 398,000 appeals, of which about 42,000 resulted in the original decisions being overturned.
These rules do not constitute a blanket ban on reverse engineering. On October 6, Anthropic expanded its Cyber Verification Program, allowing qualified information security professionals to access enhanced cybersecurity capabilities in Claude under more lenient interception classifiers. The company stated that these technologies have dual-use potential, necessitating corresponding restrictions.
OpenAI released its updated usage policy on October 21, effective November 20. The policy prohibits using its services to infringe on intellectual property and bans extracting protected internal model reasoning through unauthorized access, reverse engineering, or bypassing safeguards. This rule does not prohibit all reverse engineering of third-party software, but OpenAI's terms separately restrict reverse engineering of its own services, models, algorithms, and systems, except where prohibited by law. Violating accounts may be suspended, and users can appeal.