Securing AI
Prompt injection, jailbreaks, poisoned data and limits on AI agents.
-
US regulators start questioning banks about advanced AI models
Three US regulators, including the Federal Reserve, are asking banks about Chinese AI technology, AI agents in cybersecurity and whether they can…
-
JPMorgan, Bank of America, Citi and Schwab say outside AI agents cannot move customers’ money
Four of the largest US financial firms said in the same week that third-party AI agents may read some account data but…
-
US civilian agencies ordered to stop buying DeepSeek and remove it from their systems
A Homeland Security order bars DeepSeek from federal civilian procurement and requires agencies to remove its products, including third-party tools built on…
-
Nadella argues frontier models should be treated as insider risks
Microsoft’s chief executive published a design argument at the weekend for governing frontier models the way enterprises govern powerful insiders, including an…
-
Injected plans can slip past chain-of-thought monitors, researchers report
A new preprint reports 25 to 33 per cent monitor evasion by hiding a harmful plan in a model’s context, and finds…