Experts Refute Zuckerberg's AI Essay: Bio-Weapon Risk Claims Flawed

davidmanheim · x · 2026-08-11

In response to Meta founder Mark Zuckerberg's lengthy essay on superintelligence and AI risks, AI safety expert David Manheim shared and agreed with a detailed critique by Aaron Scher.

The critique points out that Zuckerberg's treatment of misuse risks is fundamentally flawed. Zuckerberg seems to argue that biological weapons have a favorable offense-defense balance for defenders (assuming everything will be fine if defenders get access to good models), which directly contradicts expert consensus. Furthermore, the essay's claim that "the long-term answer to risks arising from new scientific discovery is to accelerate scientific progress" is criticized as short-sighted, noting that humanity could die from an engineered pandemic before that long term arrives.

However, the critique acknowledges that the discussion on misalignment and existential risk is much better than Zuckerberg's previous statements, though it remains confused, indicating he hasn't fully followed the underlying arguments.

Original post →

More from AGI Musings

AGI Musings channel →