After the AI company Anthropic launched its new model Claude Fable5, it sparked industry controversy due to a hidden restriction measure. After facing strong opposition from the AI research community, the official announced adjustments to the relevant rules and issued an apology.

Anthropic, Claude, claude

It was reported that Claude Fable5 originally had multiple security protections, automatically switching to a lower-performing model when dealing with high-risk topics such as cybersecurity, biology, and chemistry, to avoid misuse of technology. For researchers attempting to use the model to develop competing AI, the platform previously secretly reduced the model's performance without the user's knowledge, imposing restrictions covertly. According to the service terms, the platform already prohibited users from using Claude to train similar competitive models.

This hidden restriction plan caused dissatisfaction in the industry. Many researchers believed this move was not conducive to industry development and could lead to a monopoly in the AI field. Faced with public pressure, Anthropic publicly apologized and announced the adjustment.

After the adjustment, all the platform's security measures will be fully transparent. Once the system detects a user trying to use the model to develop a high-performance competing AI, it will actively issue a reminder, clearly reject the request, or directly switch to a lower-spec model. It will no longer use the method of secretly reducing performance for control.

Key Points:

🔧 Claude Fable5 used to secretly reduce the performance of users developing competing AI, which triggered industry opposition.

🙏 Anthropic apologized for this inappropriate action and announced a comprehensive adjustment of the relevant security control rules.

👀 The new rules now include public reminders and clear rejections; all restrictions are fully visible to users.