A new AI lab wants to show its work. What should stay closed?
Nathan Lambert and Tom Zick launched Trillium Labs with plans to share AI experiments others can repeat, WIRED reports. Open work invites outside checks. Some worry the same findings could help people cause harm. Plans are not results.
What the source reports
WIRED reporter Will Knight describes a new lab started by Nathan Lambert and Tom Zick. Trillium Labs wants to publish AI experiments so others can repeat and check the work. The founders argue that openness lets more people spot errors. Others worry that high-risk findings could also be used to cause harm. The lab’s plans cover work after a model is first trained and ways it can learn from feedback. Those plans are not finished experiments or proof that openness makes AI safer. The story leaves open which findings to share, when to share them, and who should review the risks before release.
Original source
- Title
- These AI Experts Want to Do High-Stakes Research Out in the Open
- Author
- Will Knight
- Publication
- WIRED
- Date
- Friday, October 2, 2026