Anthropic Reports AI Misuse Attempts
Anthropic AI misuse Anthropic says it has detected and disrupted attempts to use its artificial intelligence models for malicious activities that could support the development of biological weapons.
The findings were detailed in the company’s latest threat intelligence report, covering misuse of its Claude AI models between December 2025 and August 2026.
Anthropic said the cases included cyber operations, surveillance, scams, fraud, influence campaigns and weapons development. The company also identified five cases involving the use of its models in ways that could potentially support biological weapons development.
Company Warns of Biological Risks
Anthropic described biological misuse as one of the most serious risks associated with frontier AI models. It warned that without appropriate safeguards, such capabilities could have catastrophic consequences.
The company also stressed that the same scientific information that could potentially be used to develop a biological weapon could have legitimate applications, including creating vaccines or treatments for diseases.
Jacob Klein, Anthropic’s head of threat intelligence, told the New York Times that the issue was highly nuanced. He said malicious actors were not necessarily approaching AI with an explicit intention to create a biological weapon, highlighting the difficulty of distinguishing legitimate scientific work from dangerous applications.
AI Models Used in Cyber Operations
Anthropic said its Claude Haiku, Sonnet and Opus models had been misused in cases that it disrupted during the reporting period.
The company said six additional cases involved Claude being used to develop software for conventional weapons, including firearms, missiles, armed drones, bombs and other munitions. The systems also included targeting and control technologies.
The report identified suspected state-sponsored groups, criminals, spyware providers, propaganda organisations and politically motivated individuals among those accused of misusing Claude.
A hacking group whose activities were consistent with Russia-based Midnight Blizzard allegedly used AI to create a system that could recognise when its malware had been detected by security software and repeatedly modify the code to evade those defences.
Hacking group ShinyHunters and China-based laboratories were also among those mentioned in the report.
Anthropic Accuses Chinese Firms Anthropic AI misuse
Anthropic also accused Chinese AI companies of attempting to reproduce Claude’s capabilities.
The company said its findings showed that cybercriminals and state-backed hackers were increasingly incorporating AI into their operations.
Anthropic said it had used the intelligence gathered from these cases to strengthen its systems and processes, with the aim of improving its ability to prevent, identify and disrupt future misuse.
It added that relevant intelligence had been shared with government authorities and industry partners when appropriate.
AI Safety Concerns Continue to Grow Anthropic AI misuse
The report comes amid growing concern among AI researchers and technology figures about the potential dangers posed by increasingly powerful AI systems.
Google recently reported that someone had attempted to use its Gemini AI tool to obtain detailed instructions for synthesising weaponised biological agents.
Anthropic’s report also comes after one of its senior safety researchers warned that there could be more than a 10% chance that AI could cause the deaths of all humans within the next decade.
OpenAI chief scientist Jakub Pachocki has separately called for the industry to consider voluntary slowdowns in AI development until stronger safeguards are established.
In the United States, Democratic Senator Bernie Sanders has proposed legislation that would temporarily halt advanced AI development and ban artificial superintelligence.
President Donald Trump has rejected some of the growing concerns about the pace of AI development. He said he was worried that the United States could be placed in a “very bad position” if it failed to win the global AI race.
The latest warnings have intensified calls for governments and technology companies to cooperate on safeguards as AI systems become increasingly capable.