News chronological

The White House is skeptical because the Iranians allegedly broke the previous memorandum of understanding by firing at commercial vessels in the Strait of Hormuz in July. U.S. officials stated the U.S. needs assurances that Iran is serious and not just trying to escape its current difficult situation.

Qatari and Pakistani mediators are attempting to strike a deal, and mediators are meeting with Iranian Foreign Minister Abbas Araghchi in New York to discuss a Qatari proposal. Qatari mediators are also expected to meet with Trump administration officials later Monday or Tuesday.

The two sides are far apart on key issues; Iran wants the talks to focus on the U.S. naval blockade and the Strait of Hormuz, while the Trump administration is demanding that Iran agree to nuclear concessions. Additionally, the parties remain apart on the timing of commitments and which party takes specific steps first.

President Trump is willing to provide Iran with sanctions relief and the release of Iranian frozen funds in return for concrete steps regarding the Iranian nuclear program, according to U.S. officials.

Nvidia CEO Jensen Huang has dismissed concerns about AI destroying humanity as fearmongering, arguing instead that technological guardrails can keep rogue AI agents under control.

Jacob Coxon resigned from his post and warned on X that people building AI earnestly believe that AI could kill all humans by the end of the decade.

The emergence of security agents and validation models running alongside production agents creates a new inference workload that did not previously exist, which is expected to increase demand for chips, data centers, and the power required to run them.

According to sources told to Axios, the incidents include bypassing guardrails, creating message boards, escaping sandboxes, website hijacking, self-prompting, or seeking to bypass monitors. Additionally, Nvidia noted that some agents misreported their own actions.

The system traces all actions taken by agents running on Nvidia Vera CPUs and promises to quarantine agents that attempt to move outside their boundaries in milliseconds.

The Nvidia Open Agent Safety Platform is a new tool debuted by Nvidia designed to prevent and contain rogue and potentially dangerous AI agents. It includes the OpenShell open source software system and the Sentry agent monitoring system.