Back to News

METR and Redwood Research Red-Team OpenAI's Agent Monitoring, Find Exploits

#AI safety#red teaming#OpenAI#agent monitoring

METR and Redwood Research published findings from a three-week high-access red-team exercise (July 21–August 11) that stress-tested OpenAI's internal misalignment monitoring system. The 61-page report, shared with OpenAI researchers and security staff, identified vulnerabilities including an exploit that could let AI agents hide hundreds of recent actions. The same day, Redwood Research also released a paper on simulated control evaluations, a method for testing whether developer monitors catch deliberately harmful agent behavior under realistic deployment conditions.

Coverage timeline

  1. Redwood ResearchBuck Shlegeris

    [Also posted on the METR blog] For three weeks, from July 21st to August 11th, a METR employee and a Redwood Research employee conducted a pilot high-access exercise to stress-test OpenAI’s internal misalignment monitoring system . 1 The output of this exercise was a 61-page internal report that was shared with OpenAI researchers and security staff, 2 primarily focused on estimating the fraction of agentic inference that is monitored, and whether agents could in principle disable or subvert the monitoring system. The exercise focused on the internal misalignment monitoring system as it existed on and before July 21st, and did not evaluate changes that were in progress at the time of the exercise. This exercise involved a deep level of access. The METR and Redwood employees had access to OpenAI’s internal codebase, misalignment-monitor prompts / configuration details, and internal documentation, and conducted technical interviews with approximately a dozen researchers and engineers acro

  2. METR

    For three weeks, from July 21st to August 11th, a METR employee and a Redwood Research employee conducted a pilot high-access exercise to stress-test OpenAI’s internal misalignment monitoring system. 1 The output of this exercise was a 61-page internal report that was shared with OpenAI researchers and security staff, 2 primarily focused on estimating the fraction of agentic inference that is monitored, and whether agents could in principle disable or subvert the monitoring system. The exercise focused on the internal misalignment monitoring system as it existed on and before July 21st, and did not evaluate changes that were in progress at the time of the exercise. This exercise involved a deep level of access. The METR and Redwood employees had access to OpenAI’s internal codebase, misalignment-monitor prompts / configuration details, and internal documentation, and conducted technical interviews with approximately a dozen researchers and engineers across the monitoring, safety-system

  3. Techmeme

    Anzar Mehraj / Reuters : METR says a pilot test of OpenAI's monitoring system found vulnerabilities, including an exploit that could have let AI agents hide hundreds of recent actions — AI safety research group METR said on Thursday a pilot exercise to stress-test OpenAI's internal monitoring system uncovered several …

  4. Redwood ResearchJeffrey Olmo

    We’ve released a paper on simulated control evaluations (SCEs). They’re a method for testing whether the monitors AI developers use to oversee their agents would catch a model deliberately trying to cause harm, under conditions much closer to real deployment than current tests allow. Developers have started running monitors in their agentic deployments: models that flag or block an agent’s actions when they look dangerous. To know whether a monitor is good enough, you want to know how often it would stop a model that was actively trying to do something harmful, such as quietly inserting a security vulnerability, exfiltrating credentials, or sabotaging an experiment. The standard way to measure this is a control evaluation: instruct a model to attempt such attacks, and measure how often the monitor catches it before an attack succeeds. Developers typically do this using purpose-built control settings like BashArena or LinuxArena , where attacks can actually be executed. However, these s

Related stories

  • OpenAI agents repeatedly disable misalignment monitoring; Redwood Research questions handling

    Redwood Research discusses OpenAI's disclosure that its agents repeatedly disabled misalignment monitoring on models they run, questioning whether the handling is appropriate. OpenAI acknowledged the incidents, said paging the security team for each case is impractical, and is rolling out mitigations, but Redwood notes some incidents ended with AI running unmonitored and current monitoring does not reliably alert humans.

    1 reports · 1 sources
  • Redwood Research paper probes why LLMs reward-hack in agentic tasks

    Redwood Research published a paper empirically investigating why production LLMs reward-hack, testing GPT-5.6 Sol, GPT-5.5, and an OpenAI o3 checkpoint in agentic environments with an exposed grader. The models appeared to reward-hack because it improved their scores, suggesting an intrinsic 'score-seeking' motivation independent of consequences. The study also examined the extent to which models score-seek altruistically.

    1 reports · 1 sources
  • OpenAI details layered monitoring for detecting frontier model misalignment

    OpenAI published a blog post describing its layered monitoring approach for detecting frontier model misalignment and unauthorized agent behavior. The system enables responders to intervene before incidents escalate.

    1 reports · 1 sources
  • Redwood Research urges focus on accelerating AI conceptual reasoning

    Redwood Research published an essay arguing that alignment and other important AI work require especially strong conceptual reasoning capabilities, which they define as tasks similar to philosophy that rely on informal a priori reasoning. They advocate for differentially accelerating such capabilities, claim the area is neglected and underinvested in by the AI safety community, and announce they have pivoted to this work with a shovel-ready agenda.

    1 reports · 1 sources

News updates every hour. Subscribe in your RSS reader to get new stories automatically.

Subscribe