Ars Technica AI

New frontier model refuses cybersecurity, biology, and chemistry queries.

Voice translations preserve speaker's tone, pacing, pitch—with SynthID watermarks for security.

Some models run in Google's cloud, but without giving Google any kind of access.

New features coming this fall alongside two-tiered, Google-powered AI model overhaul.

NotebookLM is getting a big upgrade, but it's only for AI Ultra and enterprise accounts right now.

73 packages run self-replicating stealer as soon as they're opened by an AI agent.

OpenAI to recast hit chatbot as a route to higher-margin products before a potential IPO.

Machine learning has its limits—how is it being used?

How accurate does an AI system need to be?

SpaceX won’t get easy access to billions of dollars from passive investors.

Developer felt "beaten up," with "no choice" but to shrink data center.

The Air succeeds as a minimalist, reliable fitness tracker, but Google's AI Health Coach feels unnecessary.

Robot demonstrations can distort public perceptions of robotic capabilities.

Estonian government benchmark shows how dozens of models combat Russia's "strategic narratives."

Musk can't be trusted to protect X user privacy, public commenters warn FTC.

Hyperscalers have come under scrutiny for their impact on water quality and availability.

Google must change AI Overviews after claiming users don't want "lots of sources."

Gemma 4 12B uses a new encoding scheme and token prediction to punch above its weight.

Critics say Trump plan to test AI models is short-sighted, performative.

Doubts linger over whether Meta can close the gap with rivals.

Microsoft missed the boat on apps, so get ready for agents.

International Mathematical Union endorses warning about tech industry influence.

Google's June Android feature drop includes more scam detection, more AirDrop, and yes, more AI.

Some report burning through their whole monthly "AI credit" allotment in a single day.

Pricey Instagram handles were stolen and resold before Meta patched the exploit.

Altman has an "utter disregard" for human lives, Florida AG says.

From CFD and FEA to digital twins, carmaking now involves a lot of virtualization.

Lawsuit seeks $12,000 from startup that allegedly damaged home in robot tests.

Crescent Island is an air-cooled chip that uses LPDDR5 memory.

I tried to explain OpenAI’s solution more clearly than OpenAI did.

The latest twist in paying humans to wear head cameras for robot training data.

Fine-tuning tests show "bias ... toward confidently representing the claims as true."

Undisclosed addition in jqwik instructed AI coding agents to delete app output.

As Apple tries to shrink Gemini for the iPhone, a cloud component is probably inevitable.

Here’s why Anthropic and OpenAI are on board with Illinois safety testing.

Nvidia will invest $150 billion a year to make Taiwan an AI “epicenter.”

The feds are raising the alarm about a new category of threat.

"BadHost" was found in Starlette, a package with 325 million weekly downloads.

A creepy saved post on Instagram linked man to AI porn account, FBI says.

Hugging Face debuts $2,500 bipedal robot project for builders and researchers.

Workaround flouts law that bans NTSB disclosures of cockpit audio recordings.

Trump delays AI safety testing EO, claiming it would be an innovation “blocker.”

Steven Rosenbaum explains how inaccurate quotes got into his book The Future of Truth.

SpaceX IPO filing pitches orbital data centers as Grok lags rival AI services.

Google's AI search evolution is accelerating at I/O 2026.

Figure AI's 24/7 livestream showcases human soft spot for humanoid robots.

Both tools generate hypotheses; one goes on to analyze some of the data.

AI content is getting good, but SynthID might be able to help tell truth from fiction.

Google says its more efficient Gemini 3.5 Flash is the key to your agentic AI future.

NextEra’s blockbuster deal with Dominion likely means higher bills for consumers.