Tamper-Resistance is a Moving Target We Might Not Hit

Published on June 4, 2026 1:56 PM GMTCrossposted with permission from Substack by the EA Forum Team.Subtitle: Open-Weight Model Safeguards and BioriskThe AI safety and biosecurity communities are abuzz with concerns about open-weight models. Increasingly powerful models threaten both to democratize the requisite knowledge for causing harm with biology, and to raise the ceiling of harm that a bad actor can cause, which has led to efforts to develop safeguards to prevent misuse. When models are op...