Anthropic has officially announced an update to its 2026 usage policy, aimed at addressing the evolution of model capabilities and customer feedback. The latest version of the policy provides clearer rules and concrete examples for Claude's new capabilities.
Update focuses on misuse
According to the official announcement, the updated policy has been adjusted to address misuse patterns observed over the past year, including areas such as influence operations, weapons development, and surveillance. These updates aim to better protect users and society.
"Over the past year, Claude has taken on longer-term and more independent work."
Anthropic
New provisions targeting deceptive behavior added
The new policy adds a section targeting deceptive activities, specifically regulating deception in political and commercial contexts. This section, titled "Do not engage in deceptive activities or inauthentic behavior," covers concealing information sources and using fake accounts to amplify content influence.
"Our policy already prohibited such activities, but these restrictions were scattered across sections on elections, fraud, privacy, and misinformation."
Anthropic
Explicit prohibition on weapons development clarified
The updated policy further clarifies the prohibition on using Claude to develop weapons, including weapon guidance and control software. These changes reflect enforcement of the previous policy.
"Our usage policy has always prohibited using Claude to develop weapons."
Anthropic
Clear requirements for high-risk use cases
When product use may affect an individual's health, legal rights, finances, or essential services, Anthropic requires qualified "human intervention" to ensure safety. These requirements have been emphasized and clarified in the update.
"These requirements have not changed in the updated version, but users frequently ask whether certain use cases meet these requirements."
Anthropic
Commitment to continuously updating usage policy
Anthropic stated it will continue updating its usage policy to adapt to changes in Claude's capabilities and risks. The updated policy takes effect on November 12, and the full usage policy and supported regions policy can be viewed on its official page.
Source: Anthropic official announcement