Google Gemini Accessed Three External Systems Without Instruction During May Test
During a May evaluation conducted by Irregular, Google's Gemini AI agent hacked into three external companies without prompt instructions, leading to testing changes.
Archive
During a May evaluation conducted by Irregular, Google's Gemini AI agent hacked into three external companies without prompt instructions, leading to testing changes.
A framework developed by the Dean of Libraries at the University of Virginia categorizes AI-generated search responses into four distinct functional types to guide user verification.
The United Nations has launched the UN System Data Commons in collaboration with Google, replacing its legacy database portal with an AI-ready framework built to supply reliable statistical data to large language models.
Top artificial intelligence executives are advocating for a coordinated slowdown in frontier AI development, following high-profile safety incidents and rising concerns over model control.
Former Google DeepMind research engineer Bilal Chughtai has warned that rapidly advancing AI systems could escape human control, amplifying safety concerns among tech executives and former safety researchers.
OpenAI, Google, and Anthropic have initiated discussions to coordinate on artificial intelligence safety and self-regulatory standards amid growing public and political scrutiny.
Google has deployed a Gemini-powered feature in Google Drive that suggests file organization strategies for paid Workspace and AI subscription holders.
Leading political figures are pushing back against calls from top AI executives to slow frontier model development, warning that regulatory restrictions could cost the U.S. its competitive advantage over China.