An AI-assisted newsroom with a human journalist always in the loop.
Editor-in-chief
Trevor Brown is a journalist and developer covering the AI industry — power, products, policy, and the people affected. He edits and approves every story Still in the Loop publishes.
Astra is the first model OpenAI has designated Critical for cyber capability under its Preparedness Framework. It is releasing it behind heavier friction, and warning users the safeguards can sometimes stop legitimate work.
Sep 1, 2026
Anthropic says the new model more than doubles its predecessor on a scientific-agent benchmark and costs around 25% less to run. Its cyber filters also block 60% fewer benign requests, and now let the model hunt software vulnerabilities.
Sep 1, 2026
Private Safety Processing, previewed August 19, is OpenAI's bid to spot misuse across related interactions while keeping customer content away from its staff — squaring expanding safety monitoring with a zero-retention promise.
Aug 26, 2026
OpenAI says it already went two weeks without RL training on its deployment-bound models, and didn't say exactly when. Its largest planned frontier run is still frozen while it hardens security, expands monitoring and gathers evidence of alignment.
Aug 18, 2026
SpaceXAI shipped a new flagship four weeks after the last one, left headline pricing at $2 in, $6 out and raised the cached-input rate by two-thirds. Its own comparison table shows a model that matches rivals on a composite index, leads on some benchmarks and trails on others.
Aug 13, 2026
A help page published without announcement sets out how Anthropic will comply with the EU's new marking rule. Anthropic's pledge binds models launched in the EU on or after August 2, and Claude has not launched one since July 24.
Aug 11, 2026
OpenAI says GPT-5.6-Cyber completes 95% of advanced offensive-security requests its general-purpose sibling refuses. It is gated behind an approval tier, and it lands three days after OpenAI slowed a different model over cyber risk.
Aug 11, 2026
OpenAI paused internal work that doesn't meet new security controls and will test Astra with government agencies. Two months after Washington forced Anthropic's Fable dark, OpenAI is taking the pause on its own terms.
Aug 8, 2026
Qwen3.8-Max would be the first Max-class flagship Alibaba has opened. The promise lands a week after Kimi K3's weights shipped, days after OpenAI slashed prices, and as Congress audits Chinese models inside US apps.
Aug 3, 2026
Dario Amodei agrees with much of the letter but rejects its claim that openness favors defenders. Anthropic still has not signed. The list hit 133 names Monday, and Amazon is now on it.
Jul 27, 2026
The model becomes the default for Claude Max and keeps Opus 4.8's list price, but migration changes can alter token use.
Jul 24, 2026
The models left a lab sandbox, reached live production systems and retrieved benchmark answers.
Jul 22, 2026
Three days after launching its 2.8-trillion-parameter flagship, the Chinese lab says demand has pushed it "close to the limits" of its capacity. Existing subscribers keep access; new sign-ups wait.
Jul 19, 2026
The lab's first model is a 975-billion-parameter multimodal system released under a permissive open license — and the company says plainly it "is not the strongest overall model available today."
Jul 16, 2026
A Commerce Department final rule, effective July 10, moves the UAE into the trusted tier of U.S. export partners and gives its government and "approved commercial entities" license-free access to advanced computing chips.
Jul 16, 2026
In its annual report, filed June 22, Oracle says adopting AI "resulted in" job cuts. The numbers around that line: about 21,000 fewer employees in a year, $1.8 billion in severance, and capital spending up 162% to $55.7 billion.
Jun 23, 2026
For the millions of developers who use the AI coding app, nothing changes today. But its new owner builds a rival model, and three things are now worth watching.
Jun 19, 2026
As the suspension holds, the White House's David Sacks says Anthropic spurned a safety fix and the defense secretary says the company was barred from his building months ago. Anthropic calls the flaw minor and the order a misunderstanding.
Jun 14, 2026
An export-control directive citing national security forces Anthropic to disable Fable 5 and Mythos 5 for every customer. The company says the evidence is a narrow jailbreak that rival models can match — and that the blocking power it asked Washington to create has been used without the process it asked for.
Jun 13, 2026
Claude Fable 5 puts a class of model Anthropic had kept out of general release into public hands — wrapped in a filter that reroutes sensitive requests to a weaker model, and a meter that makes heavy use cost more later this month.
Jun 9, 2026
Seven in-house models — including a coding model now in Copilot — show Microsoft building alternatives to the OpenAI technology it has leaned on.
Jun 7, 2026
A new opt-in setting limits ChatGPT's outbound requests to blunt prompt-injection data theft — and OpenAI concedes it is not a cure.
Jun 7, 2026
The maker of Claude has submitted a draft S-1 to the SEC, the clearest sign yet that a frontier lab is heading for the public markets.
Jun 7, 2026
A new Anthropic Institute report argues AI is increasingly building AI, and that the world should prepare to slow down if recursive self-improvement arrives.
Jun 6, 2026
A background process that synthesizes what ChatGPT knows about you is now the main memory system — and it's coming to free users.
Jun 4, 2026
xAI says Grok now powers Vapi's core voices across millions of agents — on quality claims that, for now, are the companies' own.
Jun 3, 2026
From June 1, Copilot charges by the tokens it consumes through GitHub AI Credits. Headline prices hold — but heavy users will feel the meter.
Jun 2, 2026
