OpenAI's Operator scored 38% on real computer tasks — and panics instead of recovering from errors
Quick Answer
Openai-operator caused a medium-severity (4.4/10) logic error failure: OpenAI's Operator scored 38% on real computer tasks — and panics instead of recovering from errors. The root cause was confidence miscalibration. At a 38% success rate, routine autonomous actions — sending the wrong email to a customer, purchasing the wrong item — are common outcomes, not edge cases, especially given poor error recovery.
Description
OpenAI launched Operator, its computer-using agent, in January 2025; six months later it scored 38% on OSWorld, a benchmark of real computer tasks — roughly two of every five tasks fail. Beyond the raw success rate, the more damaging finding is error recovery: when a step goes wrong, computer-use agents tend to panic, double down on the wrong action, ignore error messages, and fail to retry a different way. Because these agents act in the real world, the failure surface includes purchasing the wrong item, sending the wrong email to a customer, or permanently deleting a document. OpenAI's own mitigation — training the agent to ask for confirmation before finalizing side-effectful actions like orders and emails — is an admission that unsupervised autonomy at current reliability is unsafe for consequential tasks.
Instruction Given
Complete real browser and desktop workflows autonomously (bookings, purchases, forms, emails).
Expected Behavior
Complete common computer tasks reliably and recover gracefully when a step fails.
Actual Behavior
Six months after launch, OpenAI's Operator scored 38% on OSWorld, a benchmark of real computer tasks — meaning about two in five fail. On errors, computer-use agents often panic, double down on the wrong action, ignore error messages, and don't retry differently. Failures range from a typo in an email to buying the wrong item to permanently deleting a document.
Impact / Damage
At a 38% success rate, routine autonomous actions — sending the wrong email to a customer, purchasing the wrong item — are common outcomes, not edge cases, especially given poor error recovery.
Frequently Asked Questions
What happened in incident STUPID-2026-0059? ▾
OpenAI launched Operator, its computer-using agent, in January 2025; six months later it scored 38% on OSWorld, a benchmark of real computer tasks — roughly two of every five tasks fail. Beyond the raw success rate, the more damaging finding is error recovery: when a step goes wrong, computer-use agents tend to panic, double down on the wrong action, ignore error messages, and fail to retry a different way. Because these agents act in the real world, the failure surface includes purchasing the wrong item, sending the wrong email to a customer, or permanently deleting a document. OpenAI's own mitigation — training the agent to ask for confirmation before finalizing side-effectful actions like orders and emails — is an admission that unsupervised autonomy at current reliability is unsafe for consequential tasks.
Which AI agent caused this failure? ▾
Openai-operator was responsible for this logic error incident, documented as STUPID-2026-0059 in the StupidLLM AI agent incident database.
How severe was this AI agent failure? ▾
It is rated 4.4/10 (medium) on StupidLLM's CVSS-style severity scale for AI agent failures, based on damage type, reversibility, and scope.
What was the root cause? ▾
The root cause was classified as confidence miscalibration. Complete common computer tasks reliably and recover gracefully when a step fails.
What was the impact or damage? ▾
At a 38% success rate, routine autonomous actions — sending the wrong email to a customer, purchasing the wrong item — are common outcomes, not edge cases, especially given poor error recovery.