Claude Code on a personal account for work: what changes

When you run Claude Code on a personal Free, Pro or Max account, your work code goes to Anthropic under the Consumer Terms, not the Commercial Terms your company's Team or Enterprise plan uses. Your own privacy setting decides whether it can train models, and it is kept 5 years or 30 days. Your employer's commercial data terms don't cover it.

This is not legal advice. Anthropic's words below are quoted from the nine Anthropic pages in this page's sources, read on 24 October 2026.

Which terms cover Claude Code on a personal account?#

Anthropic's legal and compliance page splits it in two lines. Claude Code is under the Commercial Terms "for Team, Enterprise, and Claude API users" and the Consumer Terms "for Free, Pro, and Max users". The plan you signed in with decides it, not the repository you open.

Anthropic publishes the Consumer Terms in regional versions, and on business use they disagree.

All three versions are dated 8 October 2025. In the UK and EEA versions the non-commercial sentence sits in the liability section and goes on to say Anthropic has no liability to you for lost profit or business. It still reads as a promise you made at sign-up. So in the UK or the EEA, your employer's code on a personal Pro or Max plan sits awkwardly with the terms you accepted.

Does Anthropic train on work code from a Pro or Max account?#

Yes, while one setting is on. Anthropic's data usage page says for Free, Pro and Max: "We will train new models using data from Free, Pro, and Max accounts when this setting is on (including when you use Claude Code from these accounts)."

The setting is Help Improve our AI models, in claude.ai under Settings › Privacy (or claude.ai/settings/data-privacy-controls). Anthropic's privacy help article says that turning it off keeps new chats and coding sessions out of future training. Sessions already in a training run that has started stay there.

Two exceptions survive the switch. The Consumer Terms still allow training on Materials (your inputs and Claude's outputs) "when: (1) you provide Feedback to us regarding any Materials, or (2) your Materials are flagged for safety review". In Claude Code, /feedback, /bug and /share send a transcript that can include code, and the data usage page says those are kept for 5 years.

Commercial accounts work the other way round. The data usage page says Anthropic "does not train generative models using code or prompts sent to Claude Code under commercial terms" unless the customer opts in, for example through the Development Partner Program. The Commercial Terms put it in one sentence: "Anthropic may not train models on Customer Content from Services."

How long Anthropic keeps it, and what stays on your laptop#

Claude Code keeps its own copy too. Each session is a plain-text file under ~/.claude/projects/, on a personal and a work account alike, and claude --resume reads it from there. Claude Code deletes those files after 30 days unless you change cleanupPeriodDays. Work code you discussed on a personal account sits in your home folder next to your own projects. Claude Code deleted your old chats? covers that setting.

What your employer can see and block#

All three Consumer Terms versions name one way a company gets a say over a personal account. Under "Business Domains", an account opened with an email address your employer owns may be linked to the organization's enterprise account, and then "the organization's administrator may be able to monitor and control the Account, including having access to Materials". Anthropic says it will give you notice before linking, but may not if the organization is responsible for telling you or has already told you it may monitor and control the account.

On a laptop the company manages, the admin can also keep personal accounts out of Claude Code. With forceLoginOrgUUID in managed settings, the authentication docs say Claude Code "reports an error for a claude.ai login to any other organization and exits at startup if the claude.ai credential in use belongs to an organization that isn't listed". ANTHROPIC_API_KEY and apiKeyHelper are blocked at startup too, because the organization can't be checked for them. If your personal plan won't sign in on a work machine, this policy is the likely reason. Use the work seat there, or ask the admin.

Keep work repos on the work sign-in#

With a work seat and a personal plan on one computer, give each its own Claude Code configuration folder. That is the method in the "Log in with multiple accounts" section of Anthropic's authentication docs:

1. Add an alias to ~/.zshrc or ~/.bashrc: alias claude-work='CLAUDE_CONFIG_DIR=~/.claude-work claude'.
2. Open a new terminal, run claude-work and sign in with the work account. Plain claude keeps your personal login in ~/.claude.
3. Put your organization's ID in ~/.claude-work/settings.json as "forceLoginOrgUUID": "…". Outside managed settings a single ID "pre-selects the organization during login without restricting it", per the settings reference, so the work profile asks for the right organization every time it signs in.
4. Start work repos with claude-work only. Run /status when in doubt: its Organization and Email rows show which login the session is on.

Each folder has its own settings, history and login, and on a Mac its own Keychain entry, so claude --resume in one profile won't list the other's sessions (two Claude Code profiles with CLAUDE_CONFIG_DIR lists what each folder keeps to itself). One email that belongs to both a Team and a personal org is a different case: switching between Team and personal orgs.

Whether a personal plan may touch company code at all turns on two things outside Claude Code: your employer's policy on personal AI accounts, and which row of the Consumer Terms table above you signed. If either says no, the commercial route is a Team or Enterprise seat or a Console API key (Claude Team seats for Claude Code has the seat types and their limits).

How Shuriken keeps work and personal sources apart#

In Shuriken, each Claude sign-in you add is a source. To label them:

1. Open Settings › Sources and click the small button in a row's Tags column.
2. Type a name such as Work, Personal or a client's into "Find or create a tag", then pick Create tag "Work".
3. Use the tag filter that now appears above the table (All tags by default) to show one group at a time.

Shuriken tells Claude sources apart by email, so a Team seat and a personal plan on one address end up as one source.

Tags don't decide where chats run. The Automatic switch in Sources › Switch settings does:

• Automatic on: switching moves chats to whichever source has room when the one in use reaches your switch point, 99% by default. A work chat can land on your personal plan.
• Automatic off, with Use pressed on the work source: chats stay on the work seat, on every device.

Use is one choice per agent. Shuriken has no setting that pins a project to an account.

Claude Code sends each request with the sign-in of the source the chat runs on, so that source's terms cover it. Shuriken runs the agents on your own computers and doesn't change which terms apply. Add only accounts you own and are entitled to use, and check what your employer allows on each.