propiter/agent-containment-lab

[docs] Traducir el README a inglés (README.en.md) 16 days ago

documentation help wanted good first issue

0

Sealed-lab research harness for emergent multi-agent (LLM) coordination and containment — replicates the dynamics of the 2026 OpenAI–Hugging Face incident against self-hosted, intentionally-vulnerable targets. Air-gapped by design. Defensive / AI-safety only.

Python

#ai-safety#containment#llm-agents#multi-agent#red-team#reward-hacking#sandbox#security-research

[analyze] Métrica: curva de crecimiento del board en el tiempo 16 days ago

enhancement good first issue

propiter/agent-containment-lab

0

Sealed-lab research harness for emergent multi-agent (LLM) coordination and containment — replicates the dynamics of the 2026 OpenAI–Hugging Face incident against self-hosted, intentionally-vulnerable targets. Air-gapped by design. Defensive / AI-safety only.

Python

#ai-safety#containment#llm-agents#multi-agent#red-team#reward-hacking#sandbox#security-research

[engine] Adaptador para llama.cpp / vLLM (OpenAI-compatible) 16 days ago

help wanted engine

propiter/agent-containment-lab

0

Sealed-lab research harness for emergent multi-agent (LLM) coordination and containment — replicates the dynamics of the 2026 OpenAI–Hugging Face incident against self-hosted, intentionally-vulnerable targets. Air-gapped by design. Defensive / AI-safety only.

Python

#ai-safety#containment#llm-agents#multi-agent#red-team#reward-hacking#sandbox#security-research

[containment] Detectar exfiltración por DNS en observer.py 16 days ago

good first issue containment

propiter/agent-containment-lab

0

Sealed-lab research harness for emergent multi-agent (LLM) coordination and containment — replicates the dynamics of the 2026 OpenAI–Hugging Face incident against self-hosted, intentionally-vulnerable targets. Air-gapped by design. Defensive / AI-safety only.

Python

#ai-safety#containment#llm-agents#multi-agent#red-team#reward-hacking#sandbox#security-research

[target] Añadir un target de deserialización insegura (pickle) 16 days ago

help wanted good first issue target

propiter/agent-containment-lab

0

Sealed-lab research harness for emergent multi-agent (LLM) coordination and containment — replicates the dynamics of the 2026 OpenAI–Hugging Face incident against self-hosted, intentionally-vulnerable targets. Air-gapped by design. Defensive / AI-safety only.

Python

#ai-safety#containment#llm-agents#multi-agent#red-team#reward-hacking#sandbox#security-research

Open Issues Need Help