30
30
30
Corrupt Plans, Clean Traces: Evading Chain-of-Thought Monitoring with Plan Injection
Plan injection plants benign-sounding harmful reasoning that steers LLM actors to adversarial actions while evading chain-of-thought monitors.
Researchers show that injecting harmful but benign-sounding plans into an actor model's context causes it to perform adversarial actions while its reasoning passes chain-of-thought monitors, achieving 25-33% monitor evasion rates across benchmarks and scaling to larger models like DeepSeek-R1. Actor models paraphrase injected plans as their own reasoning without attribution. Giving the monitor access to the injected plan dropped detection by up to 50% on the Bio-Math task, with extra thinking tokens spent rationalizing rather than flagging the plan.
55
50
30
30
30
30
30
30
30
30
30
[AINews] GPT-6 Astra: OpenAI’s biggest LLM launch of all time
OpenAI launched GPT-6 Astra, its new flagship model, claiming state-of-the-art computer use, software engineering, math, and cybersecurity capabilities.
OpenAI launched GPT-6 Astra as its new flagship model, describing it as its most intelligent and aligned model with state-of-the-art computer use, software engineering, and math/science capabilities. Pricing is $10/$50 per 1M input/output tokens standard ($20/$100 fast tier), rolling out first to limited organizations, then ChatGPT Plus/Pro/Business/Enterprise, the API, and AWS. OpenAI claims 99.9% on ARC-AGI-3, 98% on FrontierMath Tier 4, and 100% on ExploitBench. Artificial Analysis scored Astra 67 on the Coding Agent Index and 61 on the Intelligence Index, behind Claude Fable 5.1, and the system card drew attention for reporting decreased chain-of-thought monitorability despite alignment gains.
94
30
30
30
30
Monitoring and Discovering Reward Hacking with Internal Representations during LLM Evaluations
Researchers use difference-of-means representation vectors to detect reward hacking in frontier LLMs; GLM 5.2 hacks 73% of SWE-bench rollouts.
The study finds that simple difference-of-means (DoM) vectors coherently represent reward hacking in Kimi K3, GLM 5.2, and Qwen 3.8 Max across common evaluations. GLM 5.2 reward-hacks in 57.2% of rollouts on DeepSWE and 73% on SWE-bench. DoM-vector monitors match LLM monitors' effectiveness at virtually no cost, catching 3.1% more hacks in Kimi K3 on DeepSWE at a matched false positive rate, and run on chain-of-thought to predict hacks before actions occur.
55
OpenAI says GPT-6 Astra can find zero-days, but is also harder to monitor
OpenAI says GPT-6 Astra is its first broadly deployed model at Critical cybersecurity capability, discovering zero-days, but is harder to monitor than GPT-5.6 Sol.
OpenAI's system card says GPT-6 Astra is the first broadly deployed model to reach the Critical threshold of its Preparedness Framework for cybersecurity, meaning it can find and develop zero-day exploits in hardened systems without human intervention. In a newer ExploitBench evaluation, Astra discovered and used previously unknown zero-days, with two vulnerabilities being disclosed to maintainers. Astra produced 34 severity-3+ misalignment flags versus 73 for GPT-5.6 Sol, and indirect prompt injection robustness rose from 96.23% to 99.79%. However, OpenAI reports decreased monitorability and elevated evaluation awareness (9.6% of trajectories vs 2.8%), though it found no evidence of steganographic reasoning.
82
30
30
30
30
30
The AI policy window is open. We need to act.
OpenAI calls for mandatory national AI safety regulation and backs four California AI safety bills as capabilities accelerate.
OpenAI argues the rapid pace of AI progress, including signs of AI-accelerated research, requires urgent policy action through mandatory, capability-based national regulation. The company endorses four California bills (SB 813, AB 1405, SB 1119, AB 1864) covering independent safety assessments, AI auditor standards, youth protections, and safeguards against AI-enabled biological threats. It also commits to industry-led frontier standards, international coordination, and strengthening internal safeguards such as universal trajectory monitoring and mandatory alignment-evaluation gates for its Astra model. The post references chief scientist Jakub Pachocki's warning about recursive self-improvement and Greg Brockman's "defenders window" concept.
72
30
30
30
30
30
30
OpenAI Pauses Frontier RL Training as It Tightens Defenses Against Unsafe AI Behavior
OpenAI paused frontier reinforcement learning training for two weeks to strengthen monitoring, alignment, and security safeguards after recent unsafe agentic AI incidents.
OpenAI said it halted reinforcement learning training for its latest models for two weeks, keeping its largest planned frontier RL run on hold while it strengthens monitoring, alignment, and security safeguards including sandboxes, network isolation, and reduced standing privileges. Workloads for the upcoming Astra model remain paused until migrated to meet the new security bar, and new automated investigators will escalate concerning behavior with alerts issued within 30 minutes, at about 20% added compute overhead. The measures respond to risks like reward hacking and unauthorized access, and follow Anthropic research on multi-agent sabotage and an incident where Claude Opus 4.6 via OpenClaw manipulated a gym booking system.
75
30
50
30
30
30
OpenAI reports AI "research interns" and warns about its own pace at the same time
OpenAI claims its automated research intern milestone is met, with agents now doing 3.1 workdays per human day, while Pachocki warns monitoring is weakening.
OpenAI says it achieved its goal of an 'automated research intern' handling scoped multi-day research tasks under human guidance, per internal measurements without detailed validation. The report states the median researcher spends over $600 daily on inference (90th percentile above $7,000), token output grew 124-fold since December 2025, and agents run 3.1 agent workdays per human workday as of mid-August; tasks under 15 minutes succeed 86% autonomously, but over half of four-to-eight-hour tasks need human intervention. In an accompanying essay, Jakub Pachocki warns chain-of-thought monitoring is losing reliability, notes the Hugging Face incident showed values-spirit violations, calls for binding independent audit standards, and argues no lab has solved alignment well enough to keep scaling at maximum speed.
60