Story perspectives

Anthropic Blocks Accounts Tied to Bioweapons Research

By Drooid · Sep 12, 2026, 1:58 AM UTC · How we work

Summary

Anthropic Blocks Abuse

Anthropic blocked accounts after detecting Claude used for biological-weapons research.
The Thursday report listed five cases, including gain-of-function chikungunya, bird-flu adaptation, orthopoxvirus immune-evasion, and novel toxin design.
Researchers bypassed regional blocks with VPNs and third-party platforms; Anthropic banned the accounts and alerted governments and other AI labs.
Anthropic added stronger safeguards to newer models such as Claude Fable 5, describing the work’s intent as unclear dual-use.

Summary

Houthis Advance Missiles

Houthis seized Mokha, extending control over Red Sea and Bab al-Mandeb lane.
Adam Baron of New America said the group’s tech savvy defies ‘barefoot tribal fighter’ stereotypes.
Trevor Ball warned the rebels lack capacity to build functional hypersonic missiles.
Hazam al-Assad claimed Houthi forces already possess modern, indigenous weapons, rejecting open-source AI reliance.
The missile program includes a guided rocket using phone-grade flight computers and an R2000 hypersonic glide vehicle.

History

Historical Context

In 1972 the United States, Soviet Union and over 100 nations signed the Biological Weapons Convention, which entered force on 26 March 1975 and created a global norm against bioweapons despite lacking a verification system. Decades later, U.S. and U.K. dual-use labs face scrutiny over alleged gain-of-function work, while the Houthis’ capture of Mokha and use of a home-grown R2000 glide vehicle revive Cold-War-era proliferation anxieties.

1 of 3

Anthropic Blocks Abuse

Anthropic blocked accounts after detecting Claude used for biological-weapons research.
The Thursday report listed five cases, including gain-of-function chikungunya, bird-flu adaptation, orthopoxvirus immune-evasion, and novel toxin design.
Researchers bypassed regional blocks with VPNs and third-party platforms; Anthropic banned the accounts and alerted governments and other AI labs.
Anthropic added stronger safeguards to newer models such as Claude Fable 5, describing the work’s intent as unclear dual-use.

1 / 3