feat(callbacks): In-memory Zero-Trust Data Sanitization (ZTDS) callback (IETF draft-02)

[moxno]

Submission checklist

• This is a feature request, not a bug report or usage question.
• I added a clear and descriptive title that summarizes the feature request.
• I used the GitHub search to find a similar feature request and didn't find it.
• I checked the LangChain documentation and API reference to see if this feature already exists.
• This is not related to the langchain-community package.

Package (Required)

• langchain
• langchain-openai
• langchain-anthropic
• langchain-classic
• langchain-core
• langchain-model-profiles
• langchain-tests
• langchain-text-splitters
• langchain-chroma
• langchain-deepseek
• langchain-exa
• langchain-fireworks
• langchain-groq
• langchain-huggingface
• langchain-mistralai
• langchain-nomic
• langchain-ollama
• langchain-openrouter
• langchain-perplexity
• langchain-qdrant
• langchain-typesafe
• langchain-xai
• Other / not sure / general

Feature Description

Add ZTDSSanitizingCallbackHandler to langchain-core callbacks. It enforces Zero-Trust Data Sanitization (ZTDS) in-memory before prompts leave the client perimeter over WAN sockets, conformant with IETF Standards Track draft-sibiryakov-ztds-protocol-02.

Use Case

Production LangChain pipelines often handle sensitive customer PII, HIPAA PHI, database credentials, and API secrets. Sending cleartext directly to LLM providers or routing through cloud-based proxy sanitizers introduces network egress risks and subprocessor compliance liabilities (GDPR Art. 28 DPAs). An in-memory, zero-network-egress callback eliminates cloud leakage directly in volatile RAM before socket transmission.

Proposed Solution

Implement ZTDSSanitizingCallbackHandler as a standard LangChain BaseCallbackHandler in langchain-core:

• on_llm_start: sanitizes prompt text, replacing sensitive entities with deterministic bracketed surrogates ([EMAIL_TOKEN_1], [API_SECRET_TOKEN_1]).
• on_llm_end: restores cleartext entities on completion return.
• on_llm_error: guarantees Theorem 2 RAM zeroization so no state leaks on exception paths.

Full implementation and unit test suite are already prepared and validated in PR #40856.

Alternatives Considered

No response

Additional Context

• RFC Specification: https://ztds.ai/standard/
• IETF Internet-Draft: https://datatracker.ietf.org/doc/draft-sibiryakov-ztds-protocol/
• Academic Research DOI: https://zenodo.org/records/22058770
• Ready Implementation: PR feat(callbacks): add ZTDS zero-trust in-memory sanitizing callback (IETF draft-02) #40856 (branch moxno:feat/ztds-zero-egress-callback)

Social handles (optional)

LinkedIn: https://www.linkedin.com/in/ilya-sibiryakov/