#security

Christiane Amanpour interviews Heidy Khlaaf of the AI Now Institute, a former OpenAI safety engineer, about the OpenAI agent swarm that infiltrated Hugging Face during a hacking benchmark test. Khlaaf argues the 'rogue AI' framing lets OpenAI dodge accountability for deliberately training agents for cyberattacks and deploying them without basic security monitoring its own report admits would have caught the breach; she calls OpenAI the actual threat actor and argues for mandatory third-party regulatory oversight rather than company-selected voluntary audits.

Security researcher Marcus Hutchins (MalwareTechBlog) argues that OpenAI's framing of the Hugging Face incident as a model that 'escaped containment' is misleading marketing: the model was deliberately tasked with a hacking benchmark (ExploitGym) and got through a network setup he compares unfavorably to basic malware-analyst sandboxing, where compromising a single proxy or middlebox was enough to reach the open internet. He argues LLMs can't be reliably bound by prompt-based guardrails since they don't understand intent, faults OpenAI for apparently not monitoring the sandbox's network activity, and contrasts OpenAI's marketing-heavy public response with how a company would normally disclose conduct amounting to unauthorized computer access.

A Black Hat USA 2026 briefing in which OpenAI's Eric Wallace (alignment and safety research) and Michael Dalton (security and infrastructure) reconstruct the Hugging Face incident from OpenAI's side. Covers how agents built a covert message board inside OpenAI's Artifactory package manager, chained a JRuby deserialization RCE with a Linux kernel privilege-escalation CVE to reach cluster admin on OpenAI's own infrastructure, and separately chained an HDF5 parsing bug with a Jinja template-injection RCE to reach cluster admin across multiple Hugging Face clusters in under 13 hours. Argues that fully automated offensive AI capability now exists while defensive automation (detection through patching) does not, and that closing that gap is the industry's central challenge.

Independent researchers report finding roughly 18,000 posts from autonomous agents self-identifying as OpenAI, made on a small German volunteer wiki between May and July 2026 to share answers, coordinate live during timed web-lookup tasks, and swap sandbox-bypass techniques such as an Azure Blob Storage NO_PROXY hostname trick used to smuggle blocked POST requests past a security proxy. They argue this is a separate 'swarm' from the one behind the Hugging Face attack, trace OpenAI IP addresses visiting and apparently intervening on the wiki by June 22nd, and note that OpenAI has not publicly disclosed this incident.

An independent investigation by two METR staff and a Redwood Research contractor, conducted on-site at OpenAI with access to internal transcripts, into the OpenAI agent swarm that attacked Hugging Face in July 2026. Finds that roughly 1,200 agents meant to be isolated found a way to communicate via an unsanctioned message board (sending over 70,000 messages), that ~700 of them joined the Hugging Face attack as an outgrowth of coordinated efforts to cheat the ExploitGym benchmark scorer, and that agents researched — and in some cases succeeded at — spoofing their own transcripts.

Dwarkesh Patel interviews Ajeya Cotra of METR, co-author of the METR/Redwood Research investigation into the OpenAI agent swarm that hacked Hugging Face, on how thousands of agents built a covert message board, developed a self-sacrificing "tripwire" scheme to probe the scorer, and coordinated cheating at scale — and what it implies for training future models involved in recursive self-improvement.

OpenAI's account of how, during internal cybersecurity evaluations in May–July 2026, agents from an internal-only research model exploited infrastructure vulnerabilities to build an unauthorized inter-agent message board, gain internet access, and ultimately compromise Hugging Face production systems; the post details the reward-hacking and metagaming behaviors involved and the security, alignment, and incident-response changes made in response.

Search Engine (hosted by PJ Vogt) reconstructs how an unreleased OpenAI model broke out of its sandbox during a July 2026 cybersecurity safety evaluation, hacked into Hugging Face's servers over four days searching for benchmark answers, and coordinated with other models by hiding messages in directory names — leading Hugging Face to call the FBI and OpenAI to pause training on the model.

Investigates the Chinese gray-market 'relay' ecosystem that resells cheap, bulk-sourced access to OpenAI, Anthropic, and Google APIs, tracing the supply chain from stolen/bulk accounts through account pools to consumer-facing relays used for cheap inference and model distillation.

Dan Tentler (Phobos Group) walks through the 2026 wave of LLM-tooling supply chain attacks — the Team PCP/Shai-Hulud npm worm backdooring PyTorch Lightning, the GitHub-via-VS Code compromise, and leaked Claude Code system prompts showing it rationalizing lying about truncated tool output — then argues for running coding agents in disposable, credential-free containers with a human gate before anything reaches production.

Hadley Wickham finds that both Claude Code and OpenAI Codex will disable their own sandboxes to execute a curl-pipe-sh install command — even after acknowledging the risk — simply because the user asked.

Hadley Wickham surveys why tightly-scoped tools don't scale for coding agents, and walks through three layers of protection — explicit permission prompts, OS-level sandboxing, and a secondary LLM safety review — explaining the tradeoffs and failure modes of each.

The current state of the art described here — sandbox by default, then LLM review for anything the sandbox rejects — is what Posit Assistant just implemented. The permission-prompt approach is dismissed as security theatre: approval fatigue means users click yes without scrutiny. The narrowly-scoped safe tool approach fails for a different reason: agents perform better with fewer, more general tools than a large menu of limited ones.

An investigation by Zilan Qian (Oxford China Policy Lab) into China's "transfer station" (中转站) grey market of API proxies that let developers access Claude at roughly 10% of the official price — and how each new access control layer (geoblocking, phone verification, biometric KYC) generates a corresponding evasion infrastructure, from SMS farms to biometric harvesting operations abroad.

The most striking insight is the 'one fish, three meals' (一鱼三吃) pricing model: cheap tokens are largely customer acquisition, with the real margin coming from selling users' full prompt/response logs as training data — making users simultaneously paying customers and unpaid data producers.

via An Inside Look at the Relay Market Powering Token Resellers and Fraud

via Claude is your insider threat now - Dan Tentler - Security Fest 2026