Google DeepMind and GovTech Singapore have built a tool that uses lightweight Gemma models running on the device to catch conversational scams in private chats. No data leaves the phone for the cloud.
Chatbot prompts show the OpenAI engineer allegedly sought help setting up a covert listening device days before one was found in a meeting room used by senior executives. The detail adds a new dimension to the arrest.
Microsoft fixed a record 1,411 vulnerabilities in its October Patch Tuesday, including a Windows flaw already used by attackers to bypass warnings on files downloaded from the internet. The company attributes the volume of findings to putting AI to work on its own code.
Cerebras CEO Andrew Feldman told TechCrunch Disrupt that AI leaders issuing doomsday warnings might want to stop talking. He argues the focus on risks is drowning out the technology's potential benefits.
OpenAI says an employee it reported to the FBI has been arrested, accused of hiding a sophisticated recording device in a meeting room and of planning to place others in sensitive areas. He was fired on October 9, the day of his arrest.
Google DeepMind has published a post on how it approached responsibility and safety in developing Gemini 4 Argon.