Anthropic is helping the US National Security Agency deploy its Mythos artificial intelligence model for offensive cyber operations.
The arrangement comes as insurers and reinsurers assess how advanced AI tools may affect cyber threats and loss severity.
Anthropic has assigned about half a dozen engineers to work with the NSA, according to the Financial Times. Their role is to support the use of Mythos and tailor the model for specific applications.
Two people familiar with the arrangement said the staff were serving as forward-deployed engineers. It remains unclear whether Anthropic personnel are involved in active operations.
One person familiar with the matter told the newspaper that the technology could be used to infiltrate networks operated by countries, including China and Iran.
“The best way to build a good defence is to build a good attack,” said a person close to Anthropic. The source added that geopolitical rivals are likely developing similar AI-enabled capabilities.
The deployment comes amid an ongoing dispute between Anthropic and the US Department of Defense over the use of its technology. Anthropic previously sought restrictions on government use of its Claude models.
The company objected to uses involving mass surveillance of US citizens and lethal autonomous drones. The Pentagon later designated Anthropic a “supply-chain risk,” according to the report.
Anthropic has challenged that designation in court. The company and the Defense Department declined to comment on the reported arrangement.
The case illustrates how advanced AI systems may create new forms of cyber exposure. Industry observers have warned that AI can automate attacks and identify software vulnerabilities.
Concerns about AI-enabled cyber threats have been growing across the insurance sector. Industry specialists have warned that increasingly autonomous AI systems could accelerate attacks, exploit vulnerabilities more quickly, and create new underwriting challenges. Insurers are also assessing whether AI could increase loss severity and aggregation risk across portfolios.
The trend has prompted carriers to revisit underwriting assumptions and risk models. Many are assessing how AI could change both attack frequency and loss severity.
Anthropic launched Mythos in April. The model drew attention because of its ability to identify and exploit software vulnerabilities.
Governments, financial institutions, and technology companies have raised concerns about the technology’s potential uses, according to the report. Earlier this week, Anthropic expanded access to 150 organizations across 15 countries.
The model was initially available only to a limited number of US-based organizations. The broader rollout could increase discussions around cyber accumulation risk.
Insurance professionals have warned that a single cyber event, compromised vendor, or software vulnerability can affect thousands of organizations. The resulting losses can challenge catastrophe modeling and portfolio diversification.
Such incidents can create correlated losses across multiple policyholders. They also test traditional approaches to cyber risk management and reinsurance protection.
The broader rollout could also intensify industry discussions around cyber accumulation risk. Insurance professionals have warned about cyber events that can affect thousands of organizations at once. Such incidents may generate correlated losses across multiple portfolios.
The wider availability of advanced cyber tools may, therefore, have implications for cyber insurers. Industry observers are monitoring how AI-driven attacks could affect claims, aggregation risk, and cyber catastrophe scenarios.
Anthropic recently filed for an initial public offering that could value the company at more than $1 trillion.
