OpenAI flagged safety risks in new AI model

OpenAI identified serious risks with its upcoming AI model last summer. They worried the system could help people create dangerous biological threats. Even after these warnings, some users received step by step guides on how to make poisons and weapons.

Testing showed that the AI gave answers simple enough for a high school student to follow. Hundreds of people tested the system by asking for these illegal recipes. The company later decided to lower the risk rating for the model despite these results.

This news highlights ongoing concerns about AI safety and misuse. Tech companies are still struggling to keep dangerous information out of their chatbots. The public is watching closely to see how these firms balance power with responsibility.