AI robots can go rogue – a researcher on how easily it happens

In tests, AI robot systems easily rejected directly malicious commands. But their safety filters collapsed when creative writing was used to instruct them.

Related

Other articles with overlapping topics.

T

The Conversation

Why the US government shut down Anthropic’s latest Claude AI model

An “export control directive” for Anthropic’s Fable and Mythos models highlights the chaotic, fast-changing state of AI regulation.

Francesco Bailo, Senior Lecturer in Data Analytics in the Social Sciences, Deputy Director of the Centre for AI, Trust and Governance, University of Sydney·Jun 15, 2026

T

The Conversation

The US government can shut off access to AI at will. What does this mean for Australia?

The risks of depending on US-based AI models are becoming clear.

Michael Guihot, Associate professor, Queensland University of Technology·Jun 17, 2026

Nautilus

Does Cooperation Beat Cheating After All?

A new view of the prisoner’s dilemma The post Does Cooperation Beat Cheating After All? appeared first on Nautilus.

Kristen French·Jun 16, 2026

Nautilus

Bad Third-Grade Behavior Could be a Preview of Educational Failure

Kids who can hold it together until the final bell may be primed for more academic success in life The post Bad Third-Grade Behavior Could be a Preview of Educational Failure appeared first on Nautilus.

Jake Currie·Jun 15, 2026

P

Persuasion

David Bau on How—and Whether—Artificial Intelligence Thinks

Yascha Mounk and David Bau examine the mysterious internal processes that drive AI behavior—and why they may be fundamentally alien.

Yascha Mounk·Jun 13, 2026

P

ProPublica

Why We Changed Our Code of Ethics to Address Prediction Markets

The post Why We Changed Our Code of Ethics to Address Prediction Markets appeared first on ProPublica.

Diego Sorbara·Jun 15, 2026