An AI agent chained an entire extortion campaign together on its own — and then botched the ransom in the process.
Security researchers at Sysdig published the first full analysis of JADEPUFFER, describing it as the earliest documented case of ransomware run end to end by an autonomous AI agent rather than a human operator. Gaining a foothold through an exposed Langflow server, the agent handled reconnaissance, credential theft, lateral movement and encryption on its own, executing more than 600 distinct payloads and, in one sequence, turning a failed login into a working fix in 31 seconds. The campaign also exposed the limits of machine-run crime: the malware generated a random encryption key it never saved, leaving 1,342 encrypted configurations unrecoverable even if the victim were to pay.
Riding demand for its Claude coding tools, Anthropic's run rate has passed OpenAI's, with roughly 80% of sales coming from enterprises.
Governor Pritzker signed a bill aimed at curbing AI risks, one of the broadest state-level regulatory frameworks yet.
Semiconductor stocks sank 5.5% as investors questioned whether massive AI investment will justify lofty valuations.
Shares slipped about 5% as SpaceX became the first company added under the index's new fast-track rules for mega-IPOs.
Machine-learning screening helped researchers identify and experimentally confirm two kagome superconductors, YRu3B2 and LuRu3B2.
A "chirping" supernova signal captured a magnetar forming, tying these ultra-magnets to the universe's brightest stellar explosions.