Repository navigation
Migrate crosswalk to OWASP Top 10 for LLM Applications 2026 - #4
Merged
Merged
Conversation
Retargets the llm-top10/ mappings from the 2025 list to the 2026 list (https://github.com/GenAI-Security-Project/GenAI-LLM-Top10/tree/main/2026/final). Eight of the ten entries changed number, and the change is a permutation, so downstream references cannot be updated with an in-order find-and-replace. MIGRATION.md carries the full map and a checklist for consumers. LLM03->LLM04 LLM04->LLM05 LLM05->LLM10 LLM06->LLM03 LLM07->LLM08 LLM08->LLM09 LLM09->LLM07 LLM10->LLM06 Renamed and re-scoped: - LLM08 Hidden Context Exposure (was System Prompt Leakage) — widened from the system prompt to all non-user-facing context, including developer instructions, retrieved policy text, and tool and function schemas - LLM10 Improper Output Handling (was Insecure Output Handling) — widened to cover insecure code generated by coding assistants at scale - LLM04 Supply Chain (was Supply Chain Vulnerabilities) - Scope notes added to LLM01 (cross-modal payloads), LLM04 (artifact provenance), and LLM05 (fine-tuning subversion) Severity re-baselined for the three entries whose rank moved, with the basis recorded in RATIONALE.md 4.5: LLM03 High->Critical, LLM06 Medium->High, LLM07 Medium->High. Also fixes three framework files that still carried entries from the 2023 list. LLM_DORA.md, LLM_FedRAMP.md, and LLM_SP800218A.md each held a stale Model DoS entry shadowing Unbounded Consumption, and Training Data Poisoning in place of Data and Model Poisoning; their quick-reference tables were keyed to pre-2025 numbering. Each was remapped by entry name, the duplicate dropped, and the entries they never had were authored: LLM10 for DORA and FedRAMP, LLM02 for SP 800-218A. The same stale cross-references in the agentic and DSGAI counterparts of those files were repaired the same way. Verified: generate.js extracts 10/10 entries from all 23 LLM framework files; validate.js reports the same 51 errors and 67 warnings as before the change (all pre-existing); 12/12 package tests pass; tsc clean. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The badge grep extracted every digit run from 'mapping%20files-67-brightgreen', so BADGE became "20\n67" and the comparison against the file total always failed. Anchor the capture on the '-<count>-' segment so the '20' inside the URL-encoded '%20' is not picked up. Pre-existing; unrelated to the 2026 migration, but it blocks this PR's gate. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Retargets the
llm-top10/mappings from the 2025 list to the 2026 final list.Breaking change
Eight of the ten entries changed number, and the change is a permutation — an in-order find-and-replace will corrupt downstream references.
MIGRATION.mdcarries the full map and a checklist for consumers.source_listis nowLLM-Top10-2026; crosswalk version2026-Q3; package4.0.0.Renamed and re-scoped
Severity re-baselining
Three entries moved with their rank. Basis recorded in
RATIONALE.md § 4.5, since the repo forbids severity changes without a cited source.Pre-existing bugs fixed along the way
LLM_DORA.md,LLM_FedRAMP.md, andLLM_SP800218A.mdstill carried entries from the 2023 list: a duplicateModel DoSshadowingUnbounded Consumption,Training Data Poisoningin place ofData and Model Poisoning, and quick-reference tables keyed to pre-2025 numbering. Each was remapped by entry name, the duplicate dropped, and the entries they never had were authored — LLM10 for DORA and FedRAMP, LLM02 for SP 800-218A. The same stale cross-references in the agentic and DSGAI counterparts were repaired the same way.Verification
generate.jsextracts 10/10 entries from all 23 LLM framework filesvalidate.jsreports the same 51 errors and 67 warnings asmain— all pre-existing, no regressiontscclean; compliance and incident reports buildKnown, left out of scope
Training Data Poisoningin four DSGAI files and one PyRIT script — a separate list's naming bugevals/garak/has no probe for LLM03 Excessive Agency (it never had one for the old LLM06)🤖 Generated with Claude Code