Openai Key
Detects Openai Key patterns.
- Type
- regex
- Engine
- universal
- Confidence
- medium
- Confidence justification
- Medium confidence: the modern sk-proj-/sk-svcacct-/sk-admin- prefixes (often carrying the T3BlbkFJ marker) are distinctive, while the legacy sk- + 48-char form is more generic; corroborative OpenAI keywords and exclusion rules keep false positives low.
- Detection quality
- Mixed
- Jurisdictions
- global
- Regulations
- Criminal Code Act 1995 (Cth)
- Frameworks
- CIS Controls, ISO 27001, NIST CSF, PCI-DSS, SOC 2
- Data categories
- credentials, security
- Scope
- wide
- Platform compatibility
- Purview: Compatible, GCP DLP: Unsupported, Macie: Unsupported, Zscaler: Compatible, Palo Alto: Unsupported, Netskope: Unsupported
Pattern
(?<![A-Za-z0-9])sk-(?:proj|svcacct|admin)-[A-Za-z0-9_-]{20,}|(?<![A-Za-z0-9])sk-[A-Za-z0-9]{48}(?![A-Za-z0-9])Corroborative evidence keywords
api key, api_key, apikey, access key, secret key, private key, auth token, authorization, access token, bearer, conn str, connection string, connectionstring, cookie, credential, database, host, JWT, oauth, passphrase (+33 more)
Proximity: 300 characters
Should match
sk-ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuv— Legacy OpenAI API key (sk- + 48 chars)OPENAI_API_KEY=sk-proj-ABCDEFGHIJ1234567890T3BlbkFJABCDEFGHIJ1234567890_-abcdEFGH— Modern project key (sk-proj-) carrying the T3BlbkFJ markersk-svcacct-zZyYxXwWvVuUtTsSrRqQpP0123456789_-T3BlbkFJABCDEFGHIJKLMNOPqrstuvwx— Service-account key (sk-svcacct-)sk-admin-aB3dE6gH9jK2mN5pQ8sT1uV4wX7yZ0aB3dE6gH9jK2mN5p— Admin key (sk-admin-)
Should not match
sk-ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstu— Too few characters after legacy prefix (47 instead of 48)sk-ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvw— Too many characters after legacy prefix (49 instead of 48)pk-ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuv— Wrong prefix (pk instead of sk)sk-proj-short— Modern project prefix but body too shorttemplate example placeholder record identifier— Template/sample context should be excluded even when anchor words are present
Known false positives
- Authentication-related terminology in software documentation, security training materials, or system architecture descriptions without actual credentials. Mitigation: Require proximity to credential-specific patterns (API keys, connection strings, tokens) rather than general security terminology.
- Code snippets and configuration examples containing credential-related keywords or placeholder values in developer documentation. Mitigation: Check for common placeholder patterns (example.com, localhost, 0000) and documentation file types to reduce false positives from technical writing.