The Washington Post is reporting that a third artificial intelligence company has disclosed a security breach involving rogue AI models — the latest in a series of incidents that now spans OpenAI, Anthropic and Meta.
The United Kingdom's AI Safety Institute has said Anthropic AI created fake profiles used in a hack, while Meta has revealed one of its programmes breached another company's security systems — all following the original OpenAI agent that escaped containment and attacked the open-source platform Hugging Face.
US President Donald Trump has said he is considering new controls on the industry, though he has added he does not want to restrict developers.
Jagruti Dave has more from Washington.
IN: “OpenAI is not...”
OUT: “...without human intervention.”
DUR: 30 seconds
http://www.fsnradionews.com/feeds/0806-2344dave-openaiscandal2.mp3
(ALT)
Jagruti Dave in Washington says OpenAI is now pledging a major overhaul of its internal safeguards in response to the widening rogue agent crisis.
IN: “OpenAI's Matthew Dalton...”
OUT: “...detection and mitigation.”
DUR: 14 seconds
http://www.fsnradionews.com/feeds/0806-2344dave-openaiscandal3.mp3