Anthropic Updates Usage Policy: Agentic Use, Cybersecurity, and Political Content
Anthropic has revised its Usage Policy effective September 15, 2025, with changes addressing agentic and cybersecurity risks, political content restrictions, law enforcement use clarity, and high-risk consumer-facing requirements. New sections explicitly prohibit malicious computer/network compromise activities while supporting legitimate security research, responding to the rapid expansion of agentic tools like Claude Code and Computer Use. The policy also narrows its previous blanket ban on political content to focus specifically on deceptive or voter-targeting uses, enabling legitimate civic and policy research. High-risk safeguards (human-in-the-loop, AI disclosure) are clarified to apply only to consumer-facing outputs, not B2B interactions.
Related guides (4)
Related events (8)
Anthropic updates Usage Policy with election integrity, high-risk use case, and privacy rules
Anthropic revised its Acceptable Use Policy (renamed Usage Policy), effective June 6, 2024, consolidating prohibited-use categories into 'Universal Usage Standards.' Key changes include explicit bans on AI-assisted election interference and political campaigning, new safety requirements for high-risk use cases (healthcare, legal), expanded access for minors via API partners with safety disclosures, and stronger privacy protections including prohibitions on biometric inference and government-directed censorship. The update reflects both evolving regulatory context and Anthropic's stated safety mission.
Anthropic publishes structured harm assessment framework covering physical, psychological, economic, and societal impacts
Anthropic has released a policy document describing their evolving framework for assessing and mitigating AI harms across five dimensions: physical, psychological, economic, societal, and individual autonomy impacts. The framework complements their existing Responsible Scaling Policy and informs decisions on usage policies, red-teaming, detection, and enforcement. Concrete examples include safeguards for computer use capabilities (fraud, phishing) and a reported 45% reduction in unnecessary refusals in Claude 3.7 Sonnet through improved handling of ambiguous prompts. Anthropic frames this as a work-in-progress and invites collaboration from the broader AI ecosystem.
Anthropic outlines election safety policies and interventions for 2024 global elections
Anthropic published a policy overview describing its three-pronged approach to election-related AI misuse in 2024: enforcing acceptable use policies that prohibit political campaigning and influence operations, red-teaming models for election-specific vulnerabilities including misinformation and voter suppression prompts, and redirecting users asking voting questions to authoritative nonpartisan sources like TurboVote and the European Parliament's elections site. The post was updated in May 2024 to cover EU users following Claude's European launch and to clarify usage policy definitions around political lobbying. The piece reflects Anthropic's cautious stance on generative AI in high-stakes civic contexts, including explicit acknowledgment of hallucination risks for real-time election information.
Anthropic August 2025 Threat Intelligence Report: Claude Misuse Case Studies
Anthropic has published its August 2025 Threat Intelligence Report documenting three real-world misuse cases involving Claude: a large-scale data extortion operation using Claude Code to automate reconnaissance and generate targeted ransom demands against 17+ organizations, a North Korean fraudulent employment scheme, and AI-assisted ransomware development by a low-skill criminal. The report highlights that agentic AI is now being weaponized for end-to-end cyberattacks rather than merely providing advisory assistance, and that AI has materially lowered the technical barrier to sophisticated cybercrime. Anthropic describes detection and countermeasures taken in each case.
Anthropic Updates Election Safeguards for Claude Ahead of 2026 US Midterms
Anthropic has published an update on its election-related safety measures for Claude, covering political bias evaluations, usage policy enforcement, and influence operation resistance testing. New model versions Claude Opus 4.7 and Sonnet 4.6 scored 95-96% on political impartiality evaluations and handled election-related policy compliance at 99.8-100% on a 600-prompt test suite. For the first time, Anthropic tested whether models can autonomously run influence operations end-to-end, finding that only Mythos Preview and Opus 4.7 completed more than half of tasks when safeguards were removed, underscoring ongoing capability concerns. Anthropic is also deploying election information banners pointing users to nonpartisan resources like TurboVote for the 2026 US midterms.
Anthropic Strengthens Regional Restrictions to Block China-Controlled Entities
Anthropic is updating its Terms of Service to prohibit access by companies whose ownership structures subject them to control from restricted jurisdictions, including China, regardless of where those companies are incorporated or operate. The new rule targets entities more than 50% owned directly or indirectly by companies headquartered in unsupported regions, closing a loophole where Chinese-controlled firms accessed Anthropic services through foreign subsidiaries. Anthropic cites national security risks including potential data sharing with intelligence services, model distillation for adversarial AI development, and support for authoritarian military objectives. The announcement also reaffirms Anthropic's advocacy for export controls, domestic AI infrastructure buildout, and national-security-focused model evaluations.
Anthropic publishes policy brief calling for targeted AI regulation within 18 months
Anthropic published a policy position paper arguing that governments have an 18-month window to enact narrowly-targeted AI regulation before risks in cyber and CBRN domains become acute. The post cites rapid capability gains—SWE-bench scores rising from 1.96% to 49% in a year, GPQA scores approaching human expert level—as evidence that frontier models are approaching meaningful misuse thresholds. Anthropic also reviews its Responsible Scaling Policy as a model for adaptive, proportionate risk governance and calls for similar frameworks to be adopted industry-wide and codified in law.
Anthropic publishes U.S. Elections Readiness summary covering policy, enforcement, and evaluation work
Anthropic released a summary of its election-integrity measures ahead of the November 5, 2024 U.S. elections, covering usage policy prohibitions on political campaigning and misinformation, automated enforcement systems, and red-teaming/vulnerability testing programs. The company implemented a TurboVote redirect for voting-information queries and released some of its automated election-safety evaluations publicly to support industry-wide efforts. The post documents Anthropic's first full election-cycle experience deploying generative AI at scale under explicit safety constraints.