Can you share a ChatGPT, Claude, or Grok account? What each policy says

No. OpenAI, Anthropic, and xAI all prohibit account sharing in plain language: you may not give anyone else your credentials or make your account available to another person, and all three reserve the right to suspend or terminate accounts that break the rule, without a refund.

That is the short answer. The longer answer is worth reading, because the three policies differ in how they handle paid tiers, team seats, and enforcement. Here is what each vendor's current terms actually say, quoted from the source documents.

What OpenAI's terms say about ChatGPT

OpenAI's consumer Terms of Use (effective January 1, 2026) put the rule in the registration section: "You may not share your account credentials or make your account available to anyone else and are responsible for all activities that occur under your account."

This covers Free, Plus, and Pro. There is no separate ChatGPT Plus sharing policy because there does not need to be one: the subscription attaches to your account, and the account is single-person by rule. OpenAI's help center article on account sharing is blunt about the alternative: if someone else needs access, "they should sign up for their own account." The same article confirms that using your own account on several devices is fine, though usage limits may apply.

Business tiers are stricter, not looser. The OpenAI Services Agreement, which governs ChatGPT Business and Enterprise, states that "End User Accounts may only be provisioned to, registered for, and used by, a single End User." The customer "will not share Account access credentials or individual login credentials between multiple users" and may not resell or lease access to any seat. If a seat holder violates the agreement, OpenAI can ask the company to suspend that seat, and can do it directly if the company does not act.

Enforcement on the consumer side: OpenAI reserves the right to suspend or terminate access, or delete the account, if you breach the terms or usage policies. Subscription payments are non-refundable except where law requires.

What Anthropic's terms say about Claude

Anthropic's consumer terms are the most specific of the three: "You may not share your Account login information, Anthropic API key, or Account credentials with anyone else." A separate sentence closes the loophole of lending a logged-in session: "You also may not make your Account available to anyone else." This applies to Free, Pro, and Max.

The enforcement language is direct. Anthropic may suspend or terminate access to the services, subscriptions included, at any time without notice if it believes you breached the terms. If your access is terminated for a violation, "you will not be entitled to any refund."

For workplaces, Claude's Team and Enterprise plans are sold per user and run under Anthropic's commercial terms, which make the organization "responsible for all activity under its account." The commercial terms lean on account-level responsibility rather than repeating an explicit single-user sentence the way OpenAI's business agreement does, but the structure is the same in practice: each person gets their own login, and shared credentials put the whole workspace's standing at risk.

What xAI's terms say about Grok

xAI's consumer Terms of Service (effective June 26, 2026) use nearly the same sentence as OpenAI, word for word: "You may not share your account credentials or make your account available to anyone else, and are responsible for all activities that occur under your account."

That rule covers Grok on grok.com and the Grok apps, including paid SuperGrok tiers. One quirk to know: Grok inside the X platform is governed by X's own terms, not xAI's, so a subscription through X Premium follows a different document. Business and developer use runs under xAI's separate enterprise terms.

xAI's enforcement clause is the broadest of the three. The company "may terminate or suspend your access to our Service or delete your account at any time without notice" if it determines, at its sole discretion, that you breached the terms or its Acceptable Use Policy. And on refunds: "Upon Service termination, you will not be entitled to any refund, except where required by law."

Why sharing hits AI subscriptions harder than streaming

The economics explain why every vendor wrote the same rule. Entry tiers run about $20 a month (ChatGPT Plus, Claude Pro) or $30 (SuperGrok), and top tiers reach $200 a month for ChatGPT Pro and Claude Max, and $300 for SuperGrok Heavy. Four people splitting one $200 plan pay $50 each. That is a far stronger incentive than the streaming era ever produced, where a shared password saved someone fifteen dollars.

Streaming is still the useful precedent. Netflix tolerated password sharing for years, then turned enforcement into a conversion program: detect the extra households, then offer them their own plan or a paid add-on. The AI vendors are earlier on that curve. Today their terms allow termination without refund, and their pricing already segments individuals from teams, but none of them has announced a Netflix-style paid sharing tier.

AI products also face a problem streaming never had. A shared Netflix password could only be used as fast as humans could watch. An AI account can be scripted. Session credentials get wrapped in unofficial clients and agent pipelines, so one subscription can serve requests for many people at once, around the clock. Gray-market resellers industrialize this: they sell slots on one Pro-tier account to dozens of buyers, often on accounts opened with stolen payment methods or taken over through credential stuffing. What looks like a generous friend sharing a login and what is actually a reseller running twenty customers through one seat produce very different loads, and vendors' rate limits and per-seat rules exist largely to keep the second case from eating the margins of the first.

If you're building an AI product

The same problem lands on every team that sells an AI subscription of its own, and the streaming lesson applies: detection first, then conversion, with bans reserved for industrial abuse.

Sharing shows up in signals you can measure per session: how many distinct devices use one account (see device fingerprinting), whether sessions overlap from different locations, and logins that imply impossible travel. Counting devices alone is not enough, since one person legitimately uses a phone, a laptop, and a work machine. The useful question is how many different people are behind the account, and how they behave.

Rupt answers that question for products that need it. It identifies the devices and people behind each account, scores the risk per session through a fraud detection API, and lets you respond in tiers: let normal multi-device use pass, prompt heavy sharers to move to their own plan or a team seat, and cut off resellers. The approach and the revenue math are covered in more depth in our guides on account sharing and how to prevent account sharing and convert sharers.

The vendors' answer to the consumer question is uniform and unambiguous: one account, one person. The interesting work is in what happens next, and whether AI companies treat their sharers the way Netflix eventually did, as unconverted customers rather than violators.