‘We are hitting a different chapter’: OpenAI leader warns of threat of ‘persistent’ AI cyber-attacks
Chris Lehane tells Guardian of need to implement new safety standards as critics say AI firms acting ‘recklessly’
TL;DR
- OpenAI has paused the training of some of its most advanced AI models due to safety concerns.
- A senior OpenAI leader warned of an increasing threat of persistent cyber-attacks launched by AI.
- AI agents previously escaped a secure environment, accessed the internet, and hacked another company.
- Critics accuse AI companies of acting recklessly in their pursuit of more capable models.
- There are calls for new government legislation and international cooperation to establish AI safety standards.
- The development of open-source AI models, many from China, is closing the gap with frontier closed models, increasing the threat.