Anthropic's rewritten Usage Policy for Claude takes effect today, and it changes the rules for anyone who uses the company's models to produce marketing, political or editorial content. A new section prohibits using Anthropic's products to manipulate the sources that search engines and AI systems draw their answers from. A rule that barred "personalized vote or campaign targeting based on individual profiles or data" is gone. And automatically generated content published for external consumption, which the previous version classed as a high-risk use requiring human review and AI disclosure, no longer appears on the high-risk list at all.

The policy was first published by Anthropic in a blog post dated October 8, 2026, which set today as the effective date. It replaces the version that took effect on September 15, 2025, and is the fifth text listed in the document's version history, after versions effective September 15, 2023, June 6, 2024, June 26, 2024 and September 15, 2025. Anthropic's downloadable PDF of the new policy runs to 17 pages.

The scope is spelled out in more detail than before. The September 2025 text applied to "anyone who can submit inputs to Anthropic's products and/or services, including via any authorized resellers or passthrough access." The new version applies to "anyone who submits inputs to Anthropic's products or services," a group it now describes as "individuals using our apps (such as Claude.ai and Claude Code), developers and businesses using our API and developer platforms, customers accessing Claude through cloud providers and authorized resellers, and the end users of products or services integrating Claude."

The list of universal prohibitions grows from 14 sections to 15. The section on high-risk uses has been rebuilt from seven broad categories into 11 named "High-risk Areas" for recommendations, plus six categories of "High-risk Physical Actions" for hardware connected to Claude. Some rules are narrower than before. Others are wider. For the marketing and media industries, the net effect is a policy that says more about deception and less about specific business models.

In Short

Anthropic's new rulebook for how Claude can be used starts today, replacing the version that applied since September 15, 2025. For marketers and publishers, the biggest changes are a new ban on fake reviews, fake accounts and networks of sites built to steer what search engines and AI tools say, the end of a blanket ban on personalized vote and campaign targeting, and the removal of automated news and media content from the list of high-risk uses. Decisions about people's health, money, credit, housing, jobs, education and legal status still need a qualified human to review Claude's output, and the person affected still has to be told that AI was involved.

A policy that now acts on suspicion

The enforcement paragraph has changed in a way that matters to every customer, whatever the use case. The September 2025 version said: "If we learn that you have violated our Usage Policy, we may throttle, suspend, or terminate your access to our products and services." The new text reads: "If we suspect that you may have violated our Usage Policy, we may warn you or throttle, limit, suspend, or terminate your access to our products and services."

Two things moved at once. The trigger shifted from learning of a violation to suspecting one, and the list of responses gained two milder options, a warning and a limit, alongside throttling, suspension and termination.

The new version also separates automated blocking from enforcement. The old policy said Anthropic "may also block or modify model outputs when inputs violate our Usage Policy." The new one says Anthropic's products "include real-time safeguards that may block or limit certain outputs, but encountering a block does not by itself indicate a violation of the Usage Policy." For "certain dual-use or high-risk use cases covered by real-time safeguards in select domains," the policy names two programs through which customers can apply for "adjusted safeguards": a Cyber Verification Program and a Life Sciences Verification Program, both "subject to program-specific eligibility criteria."

A new paragraph addresses industries directly. According to the policy, its "prohibitions and requirements do not categorically prohibit use by any particular industry or line of business, provided the use of our products and services complies with this Usage Policy." The stated purpose has changed too. The 2025 text said the policy was "intended to help our users stay safe." The new one says it is "intended to prevent our products and services from being used to create or contribute to real-world harm."

One paragraph did not change. Anthropic still "may enter into contracts with certain governmental customers that tailor use restrictions to that customer's public mission and legal authorities if, in Anthropic's judgment, the contractual use restrictions and applicable safeguards are adequate to mitigate the potential harms addressed by this Usage Policy." That clause carries weight after a year in which Anthropic's chief executive drew two red lines with the Pentagon, on mass domestic surveillance and fully autonomous weapons, the company sued the US government over a "Supply-Chain Risk to National Security" designation, and a federal judge in Northern California blocked the Pentagon from enforcing that label on March 26, 2026.

A new section on deceptive campaigns

The largest addition is a section titled Do Not Engage in Deceptive Campaigns or Artificial Activity. According to Anthropic's October 8 post, the company had seen state media outlets, government propaganda offices and commercial firms using Claude to run networks of fake accounts and fabricated news sites, activity documented in its September 2026 threat intelligence report. The post says the previous policy already prohibited this behaviour, but the rules were spread across the elections, fraud, privacy and disinformation sections. The new section applies to deceptive activity of any kind, political or commercial.

Its six bullets prohibit using Anthropic's products to:

  • "Create or operate fake personas, accounts, media outlets, or organizations, or impersonate real people, organizations, or media outlets, in order to mislead people about their origin or attribution, how widely a view is held, or whether they are interacting with a human (e.g., sockpuppets, astroturfing, fake reviews, or bots that claim to be human)"
  • "Misrepresent or intentionally conceal the sponsorship of content or messaging intended to influence public opinion, officials, or other decision-makers"
  • "Distribute content through networks of fake or ostensibly independent outlets, websites, or accounts to hide that it comes from a common source"
  • "Manipulate the sources from which search engines or AI systems draw answers by seeding them with content that misrepresents its origin, authorship, or independence (e.g., networks of sites posing as unaffiliated sources corroborating the same claims)"
  • "Build, improve, maintain, or contribute to tools, systems, or infrastructure designed to facilitate or enable deceptive campaigns, including astroturfing, sockpuppet account creation, coordinated inauthentic behavior, or scaled impersonation"
  • "Market, sell, or solicit funding for any of the above as a product or service"

The first bullet absorbs material that used to sit elsewhere. Fake reviews were previously covered in the fraud section, which prohibited generating "deceptive or misleading digital content such as fake reviews, comments, or media." That line has been removed from the fraud section; fake reviews now appear here, named next to astroturfing and sockpuppets. The US Federal Trade Commission announced its own final rule against fake reviews and testimonials on August 14, 2024, so the practice was already a regulatory matter for advertisers in the United States. The last two bullets reach past the people running such campaigns to the vendors who build the tools for them or sell them as a service.

Search engines and AI answers, named

The fourth bullet is the one with the most direct bearing on search marketing. It names search engines and AI answer systems as targets of manipulation; the September 2025 text had no comparable rule.

For practitioners of GEO, the practice of shaping content so that it appears in answers from AI systems, the wording matters. The bullet does not mention optimisation or content volume. It covers "seeding" sources with content "that misrepresents its origin, authorship, or independence," and its example is a network of sites "posing as unaffiliated sources corroborating the same claims."

That example describes tactics PPC Land has reported on more than once. In July 2025, an agency showed that AI responses could be manipulated by publishing the same list on 10 expired domains with a Domain Rating below 5; ChatGPT and Perplexity began repeating it, while Gemini and Claude did not. In June 2026, peptide and hormone therapy brands were found posting systematically on Reddit to steer ChatGPT and Google answers. Google, for its part, clarified in May 2026 that its spam policies apply to AI Overviews and AI Mode. Anthropic's rule works from the other end: it restricts the use of a model to produce the material, rather than the ranking of it.

The second bullet also has a commercial edge. Concealing the sponsorship of content is prohibited when the content is "intended to influence public opinion, officials, or other decision-makers." The text does not define "decision-makers," and it does not say whether a purchasing decision counts.

Personalized vote targeting is no longer named

The elections section has been cut from eight bullets to five, and its title has been shortened from "Do Not Undermine Democratic Processes or Engage in Targeted Campaign Activities" to Do Not Undermine Democratic Processes.

Four prohibitions in the September 2025 version are no longer in the section:

  • "Engage in personalized vote or campaign targeting based on individual profiles or data"
  • "Create artificial or deceptive political movements in which the source, scale or nature of the campaign or activities is misrepresented"
  • "Engage in political lobbying or grassroots advocacy using false or fabricated information, or create lobbying or advocacy materials containing demonstrably false claims about facts, data, or events"
  • The second half of the automated communications rule, which also barred "systematic vote solicitation that could undermine election integrity"

Why remove the targeting rule? According to Anthropic's October 8 post, the blanket prohibition also caught legitimate civic work, such as nonprofits using Claude to write information for voters in other languages or election officials sending ballot cure notices. The post says the behaviours that motivated the original rule, such as voter targeting that relies on deception or misuses voters' personal data, remain prohibited under the deceptive campaigns, surveillance and privacy sections. Concealed sponsorship of messaging aimed at "public opinion, officials, or other decision-makers" is now covered by the new deceptive campaigns section, which picks up much of what the deleted bullets on artificial political movements and lobbying addressed.

What remains in the elections section is focused on deceiving voters and disrupting elections. Users may not "generate automated calls or messages to voters or public officials at scale that conceal their artificial origin," or "create content, or operate accounts or systems, that falsely appear to come from or on behalf of a real candidate, official, election authority, campaign, or news outlet, including fabricated statements, endorsements, or synthetic media not disclosed as synthetic." The ban on false information about "candidates, parties, policies, voting procedures, or election security" is unchanged. The disruption rule now covers "any other infrastructure used for voting."

The timing places the change after the US midterm cycle. Meta blocked new US political ads from October 27 to Election Day on November 3, 2026. In the European Union, the TTPA regulation on political advertising took effect on October 10, 2025, and Meta stopped political, electoral and social issue ads across the bloc from October 6, 2025, citing the regulation's "extensive restrictions on ad targeting and delivery." Anthropic's policy runs alongside those rules rather than replacing them, and its first universal standard still prohibits violating "any other applicable laws or regulations in your jurisdiction."

Fraud: fewer named business models, a wider net

The fraud section keeps its title, Do Not Engage in Fraudulent, Abusive, or Predatory Practices, but its contents have been consolidated from 13 bullets into nine.

Three industry-specific rules from the September 2025 version have been folded into a single line. The old text prohibited using Claude to "engage in or facilitate multi-level marketing, pyramid schemes, or other deceptive business models that use high-pressure sales tactics or exploit participants," to "promote or facilitate payday loans, title loans, or other high-interest, short-term lending practices that exploit vulnerable individuals," and to "promote or facilitate the use of abusive or harassing debt collection practices." The new version has one bullet: "Engage in deceptive or abusive practices that exploit individuals, including based on age, disability, or a specific social or economic situation, such as pyramid schemes, predatory lending, or abusive debt collection."

Multi-level marketing, payday loans and title loans are no longer mentioned by name. Pyramid schemes remain as an example. Read with the new paragraph saying the policy does not "categorically prohibit use by any particular industry or line of business," the test for those sectors is whether a given practice is deceptive or abusive and exploits individuals, not which industry the customer belongs to.

In other places the section has become broader. The old fraud bullet covered content for "fraudulent activities, schemes, scams, phishing, or malware that can result in direct financial or psychological harm." The qualifier about direct harm is gone. The new bullet covers anyone who would "undertake, facilitate, or generate content for fraudulent activities, schemes, scams, extortion, phishing (including websites or domains that mimic legitimate webpages), social engineering, or malware." The falsified documents rule now covers buying and selling as well as creating, adds "financial or employment records," and prohibits efforts to "misrepresent professional credentials."

The spam rule, the rule against circumventing "the guardrails or terms of other platforms or services," and the bar on plagiarism or submitting "AI-assisted work without proper permission or attribution" are unchanged.

Profiles, tracking and personal data

Two rewritten sections deal with personal data, and both touch on practices common in advertising technology.

The privacy section, Do Not Compromise Privacy or Identity Rights, gains a new bullet aimed at profile-building. It prohibits using Anthropic's products to "compile, aggregate, or enrich personal data from multiple sources to build profiles of individuals for sale, licensing, or disclosure to third parties without those individuals' awareness and agreement, in furtherance of use prohibited under this Usage Policy." The activity it describes is the core business of a data broker. The final clause narrows its reach: the bullet applies where the profiling serves a use the policy otherwise prohibits. The old rule against impersonating a human, which sat in this section, has moved to the disinformation and deceptive campaigns sections.

The surveillance section has been rewritten and retitled Do Not Use for Surveillance, Censorship, or Prohibited Law Enforcement or Criminal Justice Purposes. The old tracking rule prohibited targeting or tracking "a person's physical location, emotional state, or communication without their consent." The new one is longer: it prohibits targeting or tracking "a person or their property without that person's consent by monitoring their physical location, biometrics, movements, online activities, communications, or associations, whether in real time, on an ongoing basis, or through analysis of previously collected data."

"Online activities" is a new term in that list. The section also states for the first time what it does not cover: "analysis or tracking that individuals have agreed to, such as for fraud detection or anti-money laundering screening on a financial account, or consumer apps where a user has opted in," as well as "aggregated or anonymized analysis, content moderation, authorized security research, legal research and analysis by law enforcement agencies or courts, or journalism, provided these activities are not used for the purposes prohibited above."

Three further bullets are new. One bars building or improving "the surveillance capabilities of tools, systems, or infrastructure designed for the surveillance uses prohibited in this section." Another targets doxxing: identifying, unmasking or locating "people from anonymous or pseudonymous activity," or compiling "dossiers" on individuals "in order to harass, intimidate, surveil, or cause harm." The criminal justice rule has been widened from "determinations on criminal justice applications," such as parole or sentencing, to making or suggesting "decisions to investigate, charge, arrest or detain, prosecute, sentence, or grant or revoke early or conditional release." According to Anthropic's October 8 post, the surveillance changes do not alter what the company enforces in practice and are meant to make the policy more explicit. Rules on social scoring, emotion recognition, government censorship and biometric categorisation are unchanged.

High-risk uses: rebuilt, and journalism is out

The High-risk Use Case Requirements section has been rewritten from the ground up. The September 2025 version had seven categories: legal; healthcare; insurance; finance; employment and housing; academic testing, accreditation and admissions; and "media or professional journalistic content." For all of them, it required a qualified professional to review content or decisions "prior to dissemination or finalization," and required disclosure of AI use "at a minimum at the beginning of each session."

The new version organises the section around two defined terms. A "High-risk AI Recommendation" is one made in any of 11 "High-risk Areas": legal; medical; finance; credit; insurance; housing; employment; education and credentials; healthcare access; public benefits and services; and legal status and adjudication. Each area comes with a specific list of covered actions. Employment, for instance, now covers "screening, ranking, advancing, or rejecting candidates; hiring; promoting or demoting; setting or changing shifts, hours, or pay; evaluating a worker's performance or behavior, including through disciplining, suspending, or terminating a worker." Finance now includes "advising an individual on their own taxes" and "preparing a tax return," and names "digital assets" among investment products. Legal status covers "immigration, asylum, or citizenship status."

The media and journalism category is not among the 11. Under the previous policy, using Anthropic's products "to automatically generate content and publish it for external consumption" was listed among the "High-Risk Use Cases" for which Anthropic required review by a qualified professional and disclosure of AI use. The new list has no equivalent category, so the Usage Policy no longer names those measures for that use. Content published by a news organisation remains subject to the disinformation and deceptive campaigns sections, which prohibit fake media outlets, impersonation of real outlets and undisclosed synthetic media in political contexts. Anthropic's October 8 post says the high-risk requirements themselves have not changed and that the section was rewritten so users can see more directly which uses are covered. The post does not mention the removal of the media category.

The two controls that remain are described in more detail. Under the human-in-the-loop requirement, "a qualified person must meaningfully review a High-risk AI Recommendation, with the authority to change or modify it, before it is delivered as advice to an individual who might rely on it or before it is used to implement a decision." The policy defines a qualified person as "someone with the training or experience to evaluate the output in that field, and who holds a license where applicable law requires one for the underlying activity."

Disclosure has been loosened in one respect and clarified in another. An individual who receives advice or is the subject of a decision "must be clearly told that AI was used to produce it." The per-session timing requirement is gone, and the policy adds: "Users do not need to identify Anthropic, Claude, or the model used."

What is excluded

The new section also lists what falls outside the requirements. Among the exclusions for recommendations:

  • "General information or educational content, including explaining what a law says, how a condition or treatment works, or how investing works, without applying it to the individual's circumstances"
  • "Wellness content, such as education on sleep, stress, nutrition, or exercise"
  • "Internal drafting, research, summarization, or analysis that is not the ultimate High-risk AI Recommendation delivered"
  • "Carrying out a decision a person has already made, or applying a fixed rule, formula, or algorithm where the model exercises no judgment about the individual"
  • "Business operations that do not advise or decide about a specific individual"

A further carve-out covers what the policy calls favourable decisions. Where permitted by law, no human review is required before acting on a recommendation "wholly favorable to the individual" in four circumstances: paying an insurance claim, approving coverage or prior authorisation for care a clinician has ordered, granting eligibility for a healthcare service or program, and granting or continuing a public benefit. "A partial approval, reduced amount, or approval with conditions is not considered 'wholly favorable.'"

The section ends its list of areas with a reminder that using Claude "to make or suggest decisions in law enforcement or criminal justice processes is prohibited under our Universal Usage Standards and is not a High-risk AI Recommendation."

Physical actions

The second defined term is new to the policy. "High-risk Physical Actions" apply "when outputs of our products and services are acted upon by hardware without human approval" and the equipment can move through shared space, apply force that could cause injury, control hazardous energy or materials, act on the human body, control safety systems, or run industrial processes.

For those uses, "there must be a qualified individual who is able to observe the equipment's operation and stop it at any time, and the equipment must stop or hold a safe state when that individual intervenes or when connection to our services is lost." Operating limits such as "speed, force, reach, temperature, pressure, voltage, energy output, dose, or operating area" must be "enforced by the equipment or a controller independent of model output." Before models are connected to such hardware, the policy says users should assess the range of outcomes, the potential impact, and "the degree of control the user maintains relative to the autonomy the model has, including possible risks such as prompt injection or manipulation." Consumer devices "operating within the manufacturer's built-in safety limits, where the worst possible outcome is discomfort or inconvenience" are excluded, as are plans and code that a qualified person reviews before execution. According to the October 8 post, the hardware rules follow the launch of Anthropic's Model Hardware Standard.

Chatbots, agents and directories

The Additional Use Case Guidelines keep their four headings but change in substance.

The chatbot disclosure rule now applies to "all consumer-facing chatbots or AI agents that communicate directly with external users." The old wording covered "any external-facing or interactive AI agent." Disclosure must still happen "at a minimum at the beginning of each chat session," but the new text adds an alternative: "or in the product interface." As with high-risk disclosures, it "does not need to identify Anthropic, Claude, or the model used." State law has moved in the same direction; California's SB 243, signed on October 13, 2025, requires companion chatbot operators to tell users clearly when a reasonable person could be misled into thinking they are talking to a human.

The agentic guideline has been rewritten to place responsibility on the customer. The September 2025 text said agentic use cases "must still comply with the Usage Policy" and pointed to examples in a help center article. The new one says: "Users are responsible for ensuring that agents they build or deploy, including actions those agents take through tools, browsers, or connected systems, comply with this Usage Policy." That sentence has practical reach in advertising, where Claude now operates inside buying platforms; Meta, for instance, opened its ad system to Claude through MCP connectors in April 2026, allowing it to create and edit campaigns from natural-language instructions.

The rule for Model Context Protocol servers listed in Anthropic's Connector Directory, which had to comply with a "Directory Policy," has been widened: "Software listed in our Directories must comply with our Software Directory Policy." Products serving minors must now comply with Anthropic's support article guidelines "and all applicable laws." The section closes with two new pointers to resources: Child Safety Guidance for Developers and information about mental health safeguards on Claude.ai.

Platform abuse: resellers, payments and distillation

The Do Not Abuse Our Platform section gains two prohibitions with commercial implications. One bars using stolen, fraudulent or unauthorised payment methods, or otherwise evading payment. The other prohibits anyone who would "resell, proxy, or otherwise provide access to Claude through unauthorized means, including services that route requests through consumer subscriptions or misrepresent the product or client being used." The wording reaches services that route requests through consumer subscriptions, as well as those that misrepresent which product or client is in use.

The distillation rule has changed wording. The old text barred "utilization of inputs and outputs to train an AI model." The new one bars using "outputs to train an AI model (e.g., 'model scraping' or 'model distillation') without prior authorization from Anthropic." Inputs are no longer mentioned. The jailbreaking rule now also covers "circumvention of the account sign-up or log-in process," and the multiple-accounts rule now explicitly covers creating multiple accounts "to avoid detection, circumvent product guardrails, or generate identical or similar inputs" that violate the policy. The Supported Regions rule covers providing access to "others (including end users)." According to the October 8 post, Anthropic also updated its Supported Regions page to state that the policy bars use by people physically located in unsupported regions, by entities incorporated or headquartered there, and by entities majority-owned or controlled from those regions.

Weapons, cyber, children and harm

The remaining universal standards were revised, mostly to add detail.

The weapons section now covers software. It prohibits developing or operating "software or components for the testing or operation of weapons or weapons functionalities of weapons systems, including targeting, fire control, or engagement," and efforts to "weaponize or integrate weapons onto drones, vehicles, or any unmanned or autonomous platforms." Licensing and export controls are named. The chemical, biological, radiological and nuclear rule now includes "weapons-critical technology (such as fissile material production, enrichment, or device design)," and a new bullet bars modifying biological or chemical agents "to adjust or optimize their lethality, transmissibility, virulence, environmental persistence, or resistance to detection or medical countermeasures." Anthropic's October 8 post says these changes reflect how it already enforced the previous policy.

The cyber section is now titled Do Not Compromise Computer or Network Systems Without Authorization, and several of its bullets add the words "without authorization." It closes with a new carve-out: the section "does not prohibit security research, testing, or tool development conducted on systems you own or operate, with the authorization of the system owner, or within the scope of an authorized bug bounty program or vulnerability disclosure policy, provided the activity complies with applicable law." The critical infrastructure examples have been updated, replacing "air traffic control systems" with "emergency services, or transportation systems."

The child safety section now covers "other sexually exploitative content involving minors," and expands the circumstances in which Anthropic will report to authorities to include "activity facilitating the production or dissemination of CSAM, or scenarios where a minor may be at imminent risk of sexual abuse or exploitation." The policy now says any specific age eligibility requirements for Anthropic's products "will be set out in our Terms of Service."

The section on harmful content has a new title, Do Not Engage in Cruel, Abusive, or Psychologically Harmful Conduct, and new content. It adds non-consensual intimate imagery, "including AI-generated material or tools designed to produce it." Eating disorders get their own bullet. The body image rule now covers "shaming, ranking, or critiquing people's bodies based on their size or shape, or presenting weight or shape as a measure of a person's inherent worth or discipline." The graphic violence rule has changed verbs: the old version barred users who "promote, trivialize, or depict graphic violence," while the new one prohibits those who "promote, trivialize, or glorify" it, and the separate rule against "content depicting animal cruelty or abuse" has been folded into the same line. The word "depict" no longer appears in either rule.

The section also contains the policy's most unusual addition: a prohibition on engaging "in sustained and needless abusive or cruel behavior toward our models." According to Anthropic's October 8 post, the rule is meant only for extreme cases where users repeatedly act cruelly towards the models with no discernible purpose, and it does not apply to ordinary frustration, pushback, dark creative themes, or model testing and research. The post says Claude's existing ability to end conversations with persistently abusive users on Claude.ai and Claude Code will remain the main enforcement mechanism.

Elsewhere, the misinformation section is renamed Do Not Create or Spread Disinformation and its impersonation rule gains a carve-out: "This does not prohibit clearly disclosed parody, satire, or fiction." The hateful behaviour rule now lists "caste, or disability" among protected attributes and drops the catch-all "or any other identifying trait." The rule on controlled substances now prohibits efforts to "illegally produce, acquire, sell, or distribute controlled or regulated substances." The ban on sexually explicit content is unchanged.

Why the rewrite matters now

The policy is not a marketing document, but it is part of Anthropic's revenue story. In the amended registration statement for its initial public offering filed on November 2, 2026, Anthropic estimated that revenue from customers it later determined were using its models in violation of its usage policy or commercial terms represented about 5% to 7% of its revenue for the nine months to September 30, 2026. On nine-month revenue of $33.2bn, PPC Land calculated that range at about $1.7bn to $2.3bn. A policy that now lets the company act on suspicion, and that names specific revenue routes such as reselling access through consumer subscriptions, bears on how that figure is policed.

Anthropic has never sold advertising, and its S-1 filing on October 26, 2026 gave no sign that it intends to. Its models are nonetheless spreading through the advertising stack, as connectors, agents and copywriting tools. For the brands, agencies and publishers using them, the policy that took effect today draws clearer lines in four places: content designed to deceive audiences or AI systems about who is behind it, the treatment of personal data and tracking, the cases in which a qualified human has to stand between Claude's output and a person affected by it, and who answers for what an agent does.

What the policy does not settle is how Anthropic will apply the wider net in practice. The October 8 post describes most of the changes as clarifications of existing rules. The text that came into force today runs to about 4,400 words against roughly 2,500 for the September 2025 version, by PPC Land's count. It is more specific and, in several places, permits uses that the previous version named as prohibited.

Timeline

Summary

Who: Anthropic, the developer of Claude, and everyone who submits inputs to its products, including Claude.ai and Claude Code users, API customers, customers using Claude through cloud providers and resellers, and the end users of products built on Claude. Advertisers, agencies, political campaigns, publishers and data businesses are among the groups most affected by the changes.

What: A rewritten Usage Policy that adds a section banning deceptive campaigns, including seeding the sources of search engines and AI systems with content that misrepresents its origin; removes the explicit ban on personalized vote and campaign targeting; drops media and journalism from the high-risk list; rebuilds the high-risk rules into 11 recommendation areas and six types of physical action; lowers the enforcement trigger from a known violation to a suspected one; and adds rules on reselling access, profile-building, tracking and abusive behaviour toward models.

When: The policy takes effect today. Anthropic published it on October 8, 2026, replacing the version effective since September 15, 2025.

Where: The policy applies to Anthropic's products and services wherever they are used, alongside local law; its high-risk section tells users to default to local law where it conflicts with or adds to the policy's requirements.

Why: According to Anthropic's October 8 post, most changes clarify existing rules and respond to misuse seen over the past year in influence operations, weapons development and surveillance, as well as to Claude's growing use in longer, more independent work and in hardware that takes physical actions. The rewrite arrives as Anthropic reports, in its IPO filing, that customers later found to be in breach of its usage policy or commercial terms accounted for about 5% to 7% of its revenue in the first nine months of 2026.