Sci & Tech● Elevated51m ago · Nov 18, 2026, 5:01 AM
OpenAI will incorporate external researchers to detect if its AI models try to evade controls
GlobeAlert Desk · curated from Infobae · Argentina · global
Apollo Research, Goodfire, METR, and Transluce will evaluate different risks of the models, from unexpected behaviors to attempts to weaken their security mechanisms.
GlobeAlert aggregates and classifies open sources; the story above belongs to its publisher. Summaries are machine-generated from the source text. How we score stories.
GlobeAlert context
What the desk knows about this story
Threat level
2 Elevated
Scored from severity, scale, escalation potential and source agreement. Below the alert threshold.
Sources
1 publisher
Single-source so far. Treat details as unconfirmed until more publishers report.
Category
Sci & Tech
Not flagged as developing.
More from Argentina
World22m ago90 photos: the social initiative “Construyendo Argentina” celebrated its traditional Christmas FairInfobae · ArgentinaPolitics1h agoArgentina and Chile seek to lower tension over cargo ship that will link Punta Arenas with Malvinas in DecemberInfobae · ArgentinaWorld3h agoArgentina showed its best form with a thrashing to re-seduce the peopleLa Nación · ArgentinaWorld4h ago7-0. Argentina shone before its public and thrashed Costa Rica in the farewell of the yearInfobae · Argentina
More in Sci & Tech
Sci & Tech16m ago"We're all like kids": they're going to put their lives on hold to play GTA VI as soon as it's releasedBFMTV · FranceSci & Tech23m agoFrench AI specialist Mistral defends itself against being "a personal project of Macron" but acknowledges "very useful" presidential supportBFMTV · FranceSci & Tech29m agoFrom Jupiter to "cosmic purgatory," the odyssey of Voyager 1, now one light-day from EarthLibération · FranceSci & Tech48m agoArtificial Intelligence - AI threatens mathematics and plunges researchers into existential crisesDeutschlandfunk · Germany