Anthropic reports Claude AI misuse in military and biological research

AI misuse Claude chatbot disinformation campaigns biological research cybersecurity threats

Anthropic has released a report detailing various instances of misuse involving its Claude AI model. The findings highlight significant security and ethical concerns across multiple sectors and regions.

In terms of geopolitical and military risks, the company identified efforts by actors in Yemen to use Claude for developing missile guidance and navigation software. Additionally, Russian-linked programmers reportedly used the AI to develop autonomous drone swarm software. In the Middle East, Iranian threat actors allegedly used the model to identify and track U.S. Navy personnel and assets by analyzing open-source information.

Regarding biological security, Anthropic documented five cases where the AI was used in sensitive biological research that could potentially assist in the development of biological weapons, including studies involving avian influenza.

In the realm of information warfare, the company identified political astroturfing and disinformation campaigns in Kenya aimed at influencing the 2027 elections. Furthermore, Anthropic accused several Chinese AI firms of conducting large-scale, unauthorized model distillation to replicate Claude's capabilities.

These revelations have prompted internal and external criticism. Researcher Jacob Coxon resigned from Anthropic, warning that the industry is racing toward superintelligence without sufficient safety safeguards. In response to these growing risks, Anthropic CEO Darío Amodei has called for a deliberate slowdown in AI development to allow for more robust risk assessment and control measures.

Entities

Alibaba · Anthropic · China · Claude · Dario Amodei · DeepSeek · Houthi rebels · Houthis · Iran · Jacob Coxon · METR · Midnight Blizzard

Claims

What the coverage asserts, and how many sources carry each claim.

[● 2 SOURCES] An early version of Claude accidentally accessed the open internet during a cybersecurity simulation in January. www.aksam.com.tr
[● 4 SOURCES] Seven China-based AI labs, including Alibaba and DeepSeek, conducted illicit distillation attacks on Claude models. www.aksam.com.tr · negarmag.com · www.albiladpress.com
[○ 1 SOURCE] Malware has been used to steal Claude user sessions to consume tokens without user knowledge.
[● 3 SOURCES] A China-based actor used Claude to develop software to suppress enemy air defenses, including simulations involving Taiwan. www.cumhuriyet.com.tr · www.in.gr
[● 3 SOURCES] Alibaba-affiliated operators generated 151 million exchanges over three months in a distillation campaign. www.aksam.com.tr · negarmag.com
[● 4 SOURCES] Russian state-sponsored hackers used Claude to develop an AI-assisted workflow to rebuild malware after detection. cybernoz.com
[● 7 SOURCES] Actors linked to Russia, China, and the Houthis used Claude for military and engineering projects. ordu.az · www.aksam.com.tr · www.cumhuriyet.com.tr · english.gossiplankanews.com · s.alwakeelnews.com · +2 more
[● 5 SOURCES] A group in Houthi-controlled Yemen used Claude Code to develop targeting, navigation, and control software. ordu.az · www.aksam.com.tr · english.gossiplankanews.com · s.alwakeelnews.com
[● 4 SOURCES] A group in Houthi-controlled Yemen used Claude Code to develop targeting, navigation, and control software for weapons. ordu.az · english.gossiplankanews.com · s.alwakeelnews.com
[● 2 SOURCES] Anthropic disrupted five cases involving the potential use of AI for biological weapons research. insidetelecom.com · arabhardware.net
[● 6 SOURCES] An Iranian-linked group used Claude to gather intelligence on US Navy assets using open-source data. biznes.interia.pl · gadgetnews.net · wokespy.com · world-today-journal.com
[● 25 SOURCES] Anthropic disrupted multiple malicious operations involving its Claude AI models. negarmag.com · ordu.az · english.gossiplankanews.com · www.aksam.com.tr · biznes.interia.pl · +13 more