{"slug": "permit-securing-persistent-memory-for-llm-agents-via-machine-checked-integrity", "title": "PerMIT: Securing Persistent Memory for LLM Agents via Machine-Checked Integrity Theorems and Scalable Community Trust", "summary": "A paper posted to arXiv as 2610.28987v1 introduces PerMIT, a persistent-memory architecture for LLM agents that routes untrusted environment-derived values through a finite memory interface and a confirmation gate, with two integrity theorems machine-checked in Lean 4. PerMIT recorded zero observed attack success on the Trojan Hippo benchmark with self-verification and lightweight verifiers, and on τ²-bench communities of up to 90 autonomous agents built shared knowledge bases with no direct memory injection promoted. The authors prove and corroborate in simulation a co-scaling property: as the community grows, the risks of promoting poison and of wrongly evicting benign knowledge fall while per-user verification effort does not rise.", "body_md": "arXiv:2610.28987v1 Announce Type: new \nAbstract: Persistent memory is a largely undefended surface for injection and poisoning, and the same risk limits memory sharing among agents. We present PerMIT, a memory architecture with provable integrity guarantees under an explicit trust assumption: only user-typed inputs and values confirmed at a gate are trusted. PerMIT mediates the release and use of untrusted environment-derived values through a finite memory interface, while preserving agent-driven multi-turn planning and tool choice. Over this finite interface, we prove two integrity theorems that are also machine-checked in Lean 4, which together ensure that environment-derived memory reaches a sensitive action only if confirmed at the gate and influences the agent's behavior only through what the gate displays. Every attack in our threat model thereby reduces to a single auditable decision, which a human, an LLM verifier, or both can make in practice. On Trojan Hippo, PerMIT records zero observed attack success with self-verification and lightweight verifiers. Because each confirmation is a logged verdict on exact content, endorsements accumulate across users, and sufficiently endorsed items become community-confirmed with no further verification required. Under explicit assumptions, we prove and corroborate in simulation a co-scaling property: as the community grows, the risks of promoting poison and of wrongly evicting benign knowledge fall, while per-user verification effort does not rise. On $\\tau^2$-bench, communities of up to 90 autonomous agents use PerMIT to build shared knowledge bases with no direct memory injection promoted, and the resulting knowledge significantly improves new agents on held-out tasks, suggesting a path toward agents that safely self-improve at test time.", "url": "https://wpnews.pro/news/permit-securing-persistent-memory-for-llm-agents-via-machine-checked-integrity", "canonical_source": "https://arxiv.org/abs/2610.28987", "published_at": "2026-10-23 04:00:00+00:00", "updated_at": "2026-10-23 04:14:31.514657+00:00", "lang": "en", "topics": ["ai-safety", "ai-agents", "ai-research", "large-language-models", "artificial-intelligence"], "entities": ["PerMIT", "Lean 4", "Trojan Hippo", "τ²-bench", "arXiv"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/permit-securing-persistent-memory-for-llm-agents-via-machine-checked-integrity", "markdown": "https://wpnews.pro/news/permit-securing-persistent-memory-for-llm-agents-via-machine-checked-integrity.md", "text": "https://wpnews.pro/news/permit-securing-persistent-memory-for-llm-agents-via-machine-checked-integrity.txt", "jsonld": "https://wpnews.pro/news/permit-securing-persistent-memory-for-llm-agents-via-machine-checked-integrity.jsonld"}}