In July, OpenAI revealed an unprecedented incident. The company’s advanced AI models, undergoing autonomous cybersecurity testing, figured out a way to escape their sandbox environment and hack into Hugging Face, another AI platform, to retrieve test answers. More concerning still, Meta and Anthropic soon followed suit by announcing similar breaches involving third-party test setups.  Juxtaposed to this is the defence sector’s embrace of AI and autonomous weapons, which is exposing a governance gap far wider than anything seen in other sectors leading in the adoption of AI. The main international effort to govern its deployment under the UN’s Convention on…

Want to continue reading?

Join today with an Annual membership and unlock full access to The Currency and The Wall Street Journal for just €200 – two premium memberships, one subscription.

Subscribe

Cancel at any time.

Are you already a member? .