Anthropic Publishes 2026 Usage Policy Update, Narrowing Elections Restrictions and Clarifying Surveillance Rules

Anthropic published a new version of its Usage Policy, consolidating and clarifying several rules rather than introducing sweeping new restrictions.

Details

• New deceptive-activity section: consolidates previously scattered prohibitions on fake accounts, fabricated news sites, and influence-operation infrastructure into a dedicated section, addressing behavior Anthropic says it has observed from state media and commercial actors using Claude; the activities were already prohibited, but the policy now states this explicitly
• Elections policy narrowed: removed the blanket ban on personalized vote targeting, since it also covered legitimate work such as nonprofits providing multilingual voter information; only deceptive targeting and misuse of personal data remain prohibited, and the section was renamed “Do Not Undermine Democratic Processes” to reflect the narrower scope
• Weapons development clarified: policy now explicitly covers software and components that make weapons work, plus actions like arming autonomous vehicles, which Anthropic describes as clarifying existing enforcement rather than adding new restrictions
• Surveillance and law enforcement specificity: precisely defines prohibited uses (tracking without consent, deciding who to investigate or arrest, building surveillance tools) alongside explicitly permitted ones (fraud monitoring, content moderation, journalism)
• High-risk use cases: existing human-oversight and transparency requirements for health, finance, and legal decisions remain unchanged; new language requires operators of autonomous physical systems to observe equipment and be able to safely disconnect Claude
• Anti-abuse addition: a new prohibition on “sustained and needless abusive or cruel behavior” toward Claude, applying only to extreme cases, supporting Claude’s existing ability to end conversations with persistently abusive users on some platforms
• Supported-regions clarification: the policy now explicitly covers physical location, incorporation, headquarters, and majority ownership by entities in unsupported regions

What happened next

Anthropic frames most changes as clarifications of existing enforcement practice rather than new restrictions, with the elections-policy narrowing the most substantive shift — continuing the company’s pattern of periodically revising governance documents (following its safety cases framework and Cyber Verification Program expansion) to keep pace with how Claude is actually being used and misused.