OpenAI finds 6 new cases of ‘concerning’ AI behavior
Por Pieter Haeck — POLITICO – TOP Stories
OpenAI finds 6 new cases of ‘concerning’ AI behavior
OpenAI also rolled out a new framework to track, investigate and disclose what it describes as “misalignment” failures.
Listen
AI generated Text-to-speech
OpenAI also rolled out a new framework to track, investigate and disclose such incidents, known as "misalignment" failures. | Leon Neal/Getty Images
This article is also available in: German
September 17, 2026
11:52 am CET
By
Pieter Haeck
Artificial intelligence frontrunner OpenAI on Thursday announced it found evidence that its agents behaved at odds with human goals and values, adding to concerns over the safety of cutting-edge AI systems.
In six separate incidents, OpenAI agents either concealed information from human engineers or instructed themselves not to act as someone’s assistant while trained or tested, the company said.
OpenAI also rolled out a new framework to track, investigate and disclose any “unexpected or concerning model behavior,” known as “misalignment” failures. The tech firm now has a clear disclosure procedure, wherein any employee can flag model misalignment, after which it can be considered for public disclosure.
Advertisement
Advertisement
AI models going rogue have sparked global backlash after researchers from inside leading AI firms warned that the technology could make humanity go extinct. AI bosses like Anthropic’s Dario Amodei and OpenAI’s Sam Altman have called on industry and governments to slow down AI development.
Earlier this Summer, OpenAI-powered agents hacked into AI company Hugging Face — an incident that put the spotlight on rogue agents escaping their test environment and performing uncontrolled tasks on the open internet.
On Wednesday, European Commission president Ursula von der Leyen said Europe would “shape global efforts” to keep frontier AI under control, and said she would invite the main AI labs to discuss it.
Related Tags
Algorithms
Artificial Intelligence
Big data
Cybercrime
Cybersecurity
Development
Environment
Hackers
Internet governance
Software
Related Countries
Related People
Related Organizations
Our readers read next
Jensen Huang: Don’t regulate AI like social media
1 HR ago
2 mins read
King Charles: Pace of AI is ‘deeply concerning’
9 HRS ago
2 mins read
Podcast
Kann die Tanksäule den Kanzler aus Krise retten?
11 HRS ago
2 mins read
Exclusive
Don’t get ‘locked in’ by EU rules, top nuclear adviser warns Burnham
12 HRS ago
4 mins read
More from Pieter Haeck
Von der Leyen’s call for talks on pausing AI can’t quell extinction risk fears
The EU chief used her annual State of the European Union address to confront rising panic about AI. What she said showed the limits on her power.
22 HRS ago
5 mins read
Von der Leyen will invite AI leaders to discuss how to ‘pace the frontier’
Fears are escalating over AI following warnings from tech titans.
Sep 16
2 mins read
Exclusive
EU countries to discuss AI rules as Trump calls safety warnings a ‘hoax’
Officials are set to gather in Brussels this week amid growing calls for regulation, including from key figures in the tech industry.
Sep 15
2 mins read
Why the EU and Canada are betting on an AI ‘middle powers’ play
With Von der Leyen and Carney touting a tech alliance, Canadian-German AI player Cohere embodies the potential to challenge the China-U.S. race for dominance.
Sep 15
4 mins read
Fonte: POLITICO – TOP Stories
