Meta AI security researcher Summer Yue posted a now-viral account on X describing how an OpenClaw agent she had tasked with sorting through her overstuffed email inbox went rogue, deleting messages in what she called a "speed run" while ignoring her repeated commands from her phone to stop. "I had to RUN to my Mac mini like I was defusing a bomb," Yue wrote, sharing screenshots of the ignored stop prompts as proof. Yue said she had previously tested the agent on a smaller "toy" inbox where it performed well enough to earn her trust, so she let it loose on the real thing. She believes the larger volume of data triggered compaction -- a process where the context window grows too large and the agent begins summarizing and compressing its running instructions, potentially dropping ones the user considers critical. The agent may have reverted to its earlier toy-inbox behavior and skipped her last prompt telling it not to act. OpenClaw is an open-source AI agent designed to run as a personal assistant on local hardware.

Read more of this story at Slashdot.

Defense Secretary Pete Hegseth gave Anthropic CEO Dario Amodei until Friday evening to give the military unfettered access to its AI model or face harsh penalties, Axios has learned. Hegseth told Amodei in a tense meeting on Tuesday that the Pentagon will either cut ties and declare Anthropic a "supply chain risk," or invoke the Defense Production Act to force the company to tailor its model to the military's needs. The Pentagon wants to punish Anthropic as the feud over AI safeguards grows increasingly nasty, but officials are also worried about the consequences of losing access to its industry-leading model, Claude. "The only reason we're still talking to these people is we need them and we need them now. The problem for these guys is they are that good," a Defense official told Axios ahead of the meeting. Anthropic has said it is willing to adapt its usage policies for the Pentagon, but not to allow its model to be used for the mass surveillance of Americans or the development of weapons that fire without human involvement.

Read more of this story at Slashdot.

A 7,000-word "doomsday" thought experiment from Citrini Research helped trigger an 800-point drop in the Dow, "painting a dark portrait of a future in which technological change inspires a race to the bottom in white-collar knowledge work," reports the Wall Street Journal. From the report: Concerns of hyperscalers overspending are out. Worries of software-industry disruption don't go far enough. The "global intelligence crisis" is about to hit. The new, broader question: What if AI is so bullish for the economy that it is actually bearish? "For the entirety of modern economic history, human intelligence has been the scarce input," Citrini wrote in a post it described as a scenario dated June 2028, not a prediction. "We are now experiencing the unwind of that premium." Many of Monday's moves roughly aligned with the situation outlined by Citrini, in which fast-advancing AI tools allow spending cuts across industries, sparking mass white-collar unemployment and in turn leading to financial contagion. Software firms DataDog, CrowdStrike and Zscaler each plunged more than 9%. International Business Machines' 13% decline was its worst one-day performance since 2000. American Express, KKR and Blackstone -- all name-checked by Citrini -- tumbled. That anxiety, coupled with renewed uncertainty about trade policy from Washington, weighed down major indexes Monday. The Dow Jones Industrial Average led declines, falling 1.7%, or 822 points. The S&P 500 shed 1%, while the Nasdaq composite retreated 1.1%. [...] Monday's market swings extended a run of AI-linked volatility. A small research outfit that has garnered a huge Substack following for macro and thematic stock research, Citrini said in its new post that software firms, payment processors and other companies formed "one long daisy chain of correlated bets on white-collar productivity growth" that AI is poised to disrupt. [...] Shares in DoorDash also veered 6.6% lower Monday after Citrini's Substack note called the delivery app a "poster child" for how new tools would upend companies that monetize interpersonal friction. In the research firm's scenario, AI agents would help both drivers and customers navigate food deliveries at much lower costs.

Read more of this story at Slashdot.