The day's verified AI news for 2026-08-10. Every claim checked against the primary source.

Meta ships a 30B agent model that fits on one gaming GPU

Meta released Muse Glimmer, a 30-billion-parameter open-weight agent model under the permissive Apache 2.0 license that shrinks to under 20 GB once quantized, small enough to run on a single 24 GB consumer graphics card.

Read on Ground Truth · primary source

OpenAI's cyber model answers 95 percent of what its flagship refuses

OpenAI expanded its Daybreak program with GPT-5.6-Cyber, a purpose-trained security model that completes 95 percent of advanced offensive-security requests where the public GPT-5.6 flagship completes about 1.5 percent.

Read on Ground Truth · primary source

Docker gives every coding agent its own microVM

Docker launched Sandboxes, a free command-line tool that runs coding agents like Claude Code and Codex inside disposable microVMs with their own kernel, filesystem, network, and private Docker engine, so a misbehaving agent cannot reach the host.

Read on Ground Truth · primary source

Sanders tells three CEOs to pause, using their own promises

Senator Bernie Sanders sent a letter on August 10 asking Sam Altman, Dario Amodei, and Mark Zuckerberg to immediately pause AI development, building his case almost entirely from the safety commitments the three companies published themselves.

Read on Ground Truth · primary source

Claude now watermarks plain text, and the EU set the date

Anthropic's support documentation now says Claude models launched in the EU on or after August 2, 2026 embed machine-readable watermarks directly in generated text, making it the first frontier lab to mark plain prose rather than only images and files.

Read on Ground Truth · primary source

NVIDIA lines up six financiers to mobilize 500 billion dollars

NVIDIA announced agreements with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to build independent financing platforms intended to mobilize more than 500 billion dollars of third-party capital for AI compute infrastructure.

Read on Ground Truth · primary source

Mistral patented letting the model write the tool call as code

Mistral AI holds a granted US patent, "Code implemented tool calls," covering an agent architecture in which a model writes a code block wrapping tool calls, a server runs it in a sandbox, pauses at each external call, and resumes with the result substituted in.

Read on Ground Truth · primary source

An AI replicated 105 ICML orals, and 34 mostly held up

The research auditing group SAI reviewed all 168 oral papers from ICML 2026, ran full execution-grounded reproductions of 105 of them, and found that only 34 reproduced more than 40 percent of the claims it attempted to check.

Read on Ground Truth · primary source

NeurIPS papers average six objective mistakes each, up from four

A study of 2,500 machine learning papers using an automated checker found that the average number of objective mistakes in a NeurIPS paper rose from 3.8 in 2021 to 5.9 in 2025, a 55 percent increase over four years.

Read on Ground Truth · primary source

The viral Riemann result an AI supposedly proved is not in the literature

A widely shared claim that Claude raised the proven fraction of Riemann zeta zeros on the critical line from 41.6 to 67.2 percent does not match any published result; the closest paper says the two-thirds figure follows only if an assumption nobody has removed can be removed.

Read on Ground Truth · primary source

You are getting this because you subscribed at groundtruth.day.