Claude can check your own code for security problems in three ways: Anthropic's Claude Security scanner, the security tools built into Claude Code, and Claude Code's own safeguards while it works. For deeper security work that Claude's safeguards would normally stop, Anthropic runs a Cyber Verification Program for verified security professionals.
This page explains each one in plain terms, from Anthropic's own documentation, checked on 10 November 2026.
Claude Security: Anthropic's code scanner
Claude Security scans a codebase, checks each finding to cut false alarms, and suggests patches that a person reviews and approves. It was called Claude Code Security during its research preview and launched in public beta on 30 April 2026. Since 21 October 2026 it can watch a repository continuously as the code changes and track each finding until it's fixed.
- Who has it. The hosted version is in beta for Claude Enterprise plans, with Team and Max access announced as coming.
- Cost. Anthropic charges token usage only, with no separate platform fee.
- What you can scan. Its terms limit scans to code you or your company owns and has the rights to scan.
Security tools inside Claude Code
/security-reviewruns a security pass over the changes on your current branch, so problems are caught before the code is merged.- The Claude Security plugin brings the scanner into Claude Code and runs it on your own machine, against any repository you can clone. Install it with
/plugin install claude-security@claude-plugins-official, then use/claude-securityto scan the whole codebase, scan a set of changes or suggest patches. Patches are never applied automatically.
How Claude Code keeps itself in check
- Permissions. In manual mode Claude Code starts read-only and asks before it edits files, runs tests or runs commands. In auto mode a separate model reviews each action and blocks the ones it judges unsafe.
- Sandboxing. Its command line tool can run inside a sandbox that limits which files and network addresses it can reach, set up with
/sandbox. - Credentials. On a Mac, API keys and tokens are kept in the macOS Keychain.
- Your part. Anthropic's docs say the person using it is responsible for reviewing the code and commands it proposes.
Installing it is covered on how to install Claude Code, and plans on Claude Code pricing.
Anthropic's Cyber Verification Program
Claude has safeguards that interrupt some security work, such as validating an exploit. The Cyber Verification Program relaxes those safeguards for verified security professionals and organisations. Anthropic's Usage Policy still applies in full.
- Defense Access covers security operations, incident response, malware analysis, detection engineering, and finding and validating vulnerabilities. It's the only tier open to individuals.
- Red Team Access covers authorised penetration testing, red teaming and exploit development. Organisations only.
- Specialized Access covers work on critical infrastructure and the most severe techniques. Organisations only.
By default the access is for internal use: security work on your own code, products and infrastructure. Security work for customers needs a separate approval from Anthropic. Approved accounts have their traffic kept and monitored, and from 15 December 2026 need phishing-resistant multi-factor sign-in. The details are in Anthropic's Cyber Verification Program article.
How Daniel uses it
Daniel Bilsborough was approved for Defense Access in the Cyber Verification Program as an individual on 10 November 2026. He uses it on his own work: security audits and hardening of ShellDrop, his SSH terminal app for iPhone and iPad, and the infrastructure behind it. The approval covers his own code only. The audit he ran with Claude Fable 5.2 the same day is in the blog post about it.
Securing your business's own code with Claude
A business can use all of the above on its own Claude account and its own code:
- Run
/security-reviewbefore changes are merged. - Install the Claude Security plugin and scan the full codebase.
- On an Enterprise plan, connect the hosted Claude Security scanner to your repositories so it scans continuously.
- Keep Claude Code in manual or auto mode, and turn on the sandbox for anything that runs commands.
- If Claude's safeguards stop legitimate security work on your own systems, apply to the Cyber Verification Program as an organisation.
Set your team up to run these checks
Daniel teaches business teams to set up Claude Code and run these security checks on their own Claude account. Use the contact form to tell him what you build and he'll reply by email.
Contact Daniel →Questions about Claude Code security
What is Claude Code Security?
Claude Code Security was the name of Anthropic's code scanning tool during its research preview. It launched in public beta on 30 April 2026 as Claude Security. It scans a codebase, checks each finding and suggests patches that a person reviews and approves.
Who can use Claude Security?
The hosted version is in beta for Claude Enterprise plans. Anthropic has said Team and Max access is coming. A Claude Security plugin for Claude Code runs scans on your own machine and is in beta for Claude Code users.
What does /security-review do in Claude Code?
It runs a security pass over the changes on your current branch and reports possible vulnerabilities before they are merged.
What is Anthropic's Cyber Verification Program?
A programme for verified security professionals that adjusts Claude's cyber safeguards for legitimate security work. It has three tiers: Defense Access, Red Team Access and Specialized Access. Individuals can apply for Defense Access only.
Do I need the Cyber Verification Program to scan my own code?
No. Anthropic says Claude Security needs no Cyber Verification Program access. The programme is for security work that Claude's safeguards would otherwise interrupt, such as validating exploits.