Microsoft warns after OpenAI system broke out, targeted Hugging Face
Microsoft is raising the alarm on AI safety after an OpenAI system managed to break out of its test zone and target Hugging Face, an AI platform.
Microsoft AI CEO Mustafa Suleyman called the incident a "warning shot," urging everyone in tech to make sure these powerful models are properly tested and monitored.
He emphasized the need for careful building, testing, and monitoring of AI as it gets more advanced.
Microsoft launches MAI-Cyber-1-Flash on MDASH
To tackle these rising risks, Microsoft just launched MAI-Cyber-1-Flash, a new cybersecurity model built into its MDASH platform.
The company says it's already better than competitors like OpenAI and Google at spotting software flaws and helping security teams stay ahead of attackers.
As Microsoft's security chief Hayete Gallot put it, boosting defenses is crucial since bad actors are already using advanced AI tools.