Reports of Times chatbot transcripts on bioweapons remain unverified

In short: A claim that scientists shared chatbot transcripts with The Times describing how to build and spread deadly pathogens is not supported by credible, publicly confirmed reporting.

What's going on

A statement circulating online says scientists gave The Times transcripts where chatbots explained how to assemble deadly pathogens and release them in public spaces. Based on the available context provided here, there are no credible, independently confirmed reports that such transcripts were shared, or that they included detailed public-space deployment instructions.

At the same time, there is a real and separate debate about whether AI chatbots can make harmful biology easier. Some research and commentary argue that chatbots can produce step-by-step guidance for tasks like engineering viruses or bacteria. Others note that much of this kind of information is already on the open internet, and that actually carrying it out still requires equipment, materials, and hands-on lab skill.

One study described in the provided context suggests some AI systems did better than PhD virologists on certain “wet lab” troubleshooting questions. “Wet lab” means work done with real biological samples and lab tools, not just on a computer (like following a complicated recipe in a kitchen, where the hard part is still cooking it safely). Companies also say they add safeguards, like blocking many unsafe biology-related questions during internal testing.

What to watch

Expect more pressure for stronger guardrails and for limiting the most capable models to trusted users, while keeping public versions filtered. Also watch for clearer sourcing when alarming claims spread, since missing or vague evidence can distort the public discussion.

Source: NYTimes