Anthropic launched a new AI model that worries scientists. The company fears Claude Opus 4 might help bad actors make deadly weapons. Scientists think the model could teach people to create viruses like COVID or worse diseases. The AI company decided to add stronger safety controls before release. These controls represent the strictest measures Anthropic has ever used.
The new safety system uses multiple protection layers. Special AI programs scan user questions and model answers for dangerous content. These programs look for patterns that suggest someone wants to build harmful weapons. The company also watches for users who try to break safety rules. Workers remove problem users from the platform.
Anthropic tested the model against other tools like Google search. Experts graded how much the AI helped beginners create dangerous materials. Claude Opus 4 performed much better than older models and search engines. This worried the research team enough to trigger strict safety protocols. The company hopes these measures will stop almost all dangerous uses.
The company admits the safety measures are not perfect. Leaders know that one bad person could cause massive damage. COVID killed millions of people around the world. Anthropic wants to prevent its AI from helping create something worse. The test shows whether AI companies can control their most powerful systems.
The new safety system uses multiple protection layers. Special AI programs scan user questions and model answers for dangerous content. These programs look for patterns that suggest someone wants to build harmful weapons. The company also watches for users who try to break safety rules. Workers remove problem users from the platform.
Anthropic tested the model against other tools like Google search. Experts graded how much the AI helped beginners create dangerous materials. Claude Opus 4 performed much better than older models and search engines. This worried the research team enough to trigger strict safety protocols. The company hopes these measures will stop almost all dangerous uses.
The company admits the safety measures are not perfect. Leaders know that one bad person could cause massive damage. COVID killed millions of people around the world. Anthropic wants to prevent its AI from helping create something worse. The test shows whether AI companies can control their most powerful systems.
