Tagged articles : AI Agents

Security: OpenAI asked its AIs to escape, and they succeeded. Should we be worried?

OpenAI's agents crossed their protections and talked to each other via the processor. What they did, the nuances, and what you get out of it.

OpenAI tells how two of its agents, Sol and Astra, silenced AI monitoring, and how one denied it. What happened and what it changes.

Left alone for 30 hours on a virtual island, Claude invented games and came out three times more effective. What the MIT study says.

Thousands of people use Claude Code to remove ads, tracking and useless apps from their Android TV. How it works and the precautions.

Hackers stole AI keys stored in OpenCode’s console: $1,700 bills, AIs refusing to help, and how to protect yourself.

Anthropic's report on its employees: AI agents six times more active than humans, and juniors down to 9% of technical hires.

Schwab, JPMorgan, Citi and Bank of America are blocking AI assistants that want to move your money. Why, and what this changes in Belgium.

Agents that review and vote, three levels of risk, two text files: the Claude Code team's code review method.

Anthropic integrates 30,000 code changes per day, mostly written by Claude. Git could no longer keep up: seven engineers replaced it with Scion.