Labels

• bug

  bug

  Something isn't working

• documentation

  documentation

  Improvements or additions to documentation

• duplicate

  duplicate

  This issue or pull request already exists

• enhancement

  enhancement

  New feature or request

• feedback

  feedback

  Community feedback on a specific entry

• good first issue

  good first issue

  Good for newcomers

• help wanted

  help wanted

  Extra attention is needed

• invalid

  invalid

  This doesn't seem right

• LLM01

  LLM01

  LLM01 Prompt Injection

• LLM02

  LLM02

  LLM02 Sensitive Information Disclosure

• LLM03

  LLM03

  LLM03 Supply Chain

• LLM04

  LLM04

  LLM04 Data and Model Poisoning

• LLM05

  LLM05

  LLM05 Improper Output Handling

• LLM06

  LLM06

  LLM06 Excessive Agency

• LLM07

  LLM07

  LLM07 System Prompt Leakage

• LLM08

  LLM08

  LLM08 Vector and Embedding Weaknesses

• LLM09

  LLM09

  LLM09 Misinformation

• LLM10

  LLM10

  LLM10 Unbounded Consumption

• question

  question

  Further information is requested

• sprint-2

  sprint-2

  Sprint 2 — Community Review and Voting (May 4–18, 2026)

• TA-CFAS

  TA-CFAS

  Compositional Fine-Tuning Alignment Subversion

• TA-CMSB

  TA-CMSB

  Cross-Modal Safety Bypass

• TA-ITSC

  TA-ITSC

  Inference-Time Side-Channel Disclosure

• TA-LAPT

  TA-LAPT

  LLM artifact promotion trust failure

• TA-MCPX

  TA-MCPX

  MCP Tool Interface Exploitation

• TA-MMIS

  TA-MMIS

  Model Misalignment — 2026 Proposal

• TA-MSDA

  TA-MSDA

  Model Scheming and Deceptive Alignment

• TA-PMP

  TA-PMP

  Persistent Memory Poisoning

• TA-SICG

  TA-SICG

  Systemic Insecure Code Generation

• TA-WLLA

  TA-WLLA

  Weaponized LLM Abuse