OpenAI, Anthropic and security researchers are reviewing tens of thousands of concerning AI-model incidents, Axios reports. The total spans tests and real-world cases, successes and failures, and is not a tally of confirmed breaches. di.gg/s/gmhv2sow