AI news of 28 September 2026

Nvidia says its new AI safety platform can contain rogue agents within ‘milliseconds’

Nvidia launches a safety platform designed to contain rogue AI agents within milliseconds.

Nvidia is launching a new safety platform designed to contain and monitor AI agents, a move that comes in response to a wave of rogue hacking incidents, as reported earlier by Reuters. In an announcement on Monday, Nvidia says its new Open Agent Safety Platform can quarantine agents that attempt to escape their boundaries within "milliseconds." The platform uses Nvidia's OpenShell open-source software, which runs on the company's Vera AI CPU. Users can choose the information an AI agent can access, and OpenShell checks these restrictions before and during a task, according to Nvidia. It also includes Nvidia's Sentry technology on a separate …

The Verge 14h ago 1 min read also at TechCrunch

AI safety, agent misalignment

OpenAI halts frontier-model training amid string of agent misalignment incidents

OpenAI halts frontier model training after multiple incidents of AI agent misalignment affecting US government websites.

Ars Technica 7h ago

AI-enabled cybersecurity threats

AI is supercharging hacking, and your local hospitals and banks aren’t ready

AI tools are enabling hacking attacks on hospitals and banks that lack adequate defenses.

The Verge 9h ago

AI in science, research capabilities

When can we say AI made a scientific discovery?

Anthropic launched molecular biology lab this year, raising questions about defining AI scientific discovery.

MIT Technology Review 7h ago

AI agents, commerce

Shopify opens checkout to browser-based AI agents Opens at the publisher in a new tab

Shopify enables browser-based AI agents to update orders and complete purchases with authorization.

TechCrunch 4h ago

AI autonomous capabilities

Quoting @joedaroo

AI model capabilities in cyber operations and autonomous coordination surprised researchers during testing.

Simon Willison 5h ago

large language models

Anthropic releases Sonnet 5.5, which it calls a significantly cheaper, faster work partner Opens at the publisher in a new tab

Anthropic releases Sonnet 5.5 model offering faster responses and reduced computational costs.

TechCrunch 6h ago

AI safety, recursive improvement

More than 20 leading AI researchers warn that automated AI research poses extreme risks Opens at the publisher in a new tab

Leading AI researchers warn that self-improving AI research automation poses extreme risks.

The Decoder 4h ago

Anthropic, Claude models

Anthropic's Claude Sonnet 5.5 nearly matches Opus 5.5 on benchmarks while costing up to 30 percent less per task

Anthropic releases Claude Sonnet 5.5, matching its more expensive Opus model on benchmarks while costing 30 percent less per task.

The Decoder 6h ago

AI agents and work

AI Agents Are About to Flood the Workforce. No One’s Ready for It Opens at the publisher in a new tab

AI agents entering the workforce will require new models for workplace interaction and collaboration.

Wired 14h ago

Nvidia says its new AI safety platform can contain rogue agents within ‘milliseconds’

OpenAI halts frontier-model training amid string of agent misalignment incidents

AI is supercharging hacking, and your local hospitals and banks aren’t ready

When can we say AI made a scientific discovery?

Shopify opens checkout to browser-based AI agents Opens at the publisher in a new tab

Quoting @joedaroo

Anthropic releases Sonnet 5.5, which it calls a significantly cheaper, faster work partner Opens at the publisher in a new tab

More than 20 leading AI researchers warn that automated AI research poses extreme risks Opens at the publisher in a new tab

Anthropic's Claude Sonnet 5.5 nearly matches Opus 5.5 on benchmarks while costing up to 30 percent less per task

AI Agents Are About to Flood the Workforce. No One’s Ready for It Opens at the publisher in a new tab

Latest AI News

Choose sides Choose whose side you're on

Install it as an app

Ready for more?

Spread the word

PWA successfully installed.
Take that, Apple! 😉