Members read every story in the archive, re-verified. Sources and corrections stay open to everyone.
What the pen marks mean
rememberThe takeaway: what to remember
figureThe number that matters
evidenceThe evidence line
rulingThe ruling
claimThe claim that fails
Each mark is drawn once, as you reach it. Numbers are only circled when they come from the story’s own figures.
01 / SafetyBS
Nvidia says its new agent safety platform could have prevented the Hugging Face hack. No test against that hack is on the record.
Nvidia (an unnamed company representative on a press call, as reported by CNBC) ·
Nvidia launched the Open Agent Safety Platform, built on OpenShell software and the Sentry watchdog. CNBC reports that an unnamed Nvidia representative told reporters on a Sunday call that the platform could have prevented OpenAI's Hugging Face incident in July.
Members · 30 days free
See what the evidence actually shows.
Members read the full check on every story: what the evidence shows, why it matters to you and the one line to take with you. Every past edition, re-verified, and the Receipts Pack come with it. A$89 a year, about A$0.24 a day.
First membership: 30 days free, then A$89 a year. One introductory trial per customer. Card required; renews annually until cancelled. Cancel before the trial ends to avoid the first charge.
Open the evidence4 source pages +
The claim we checked
Nvidia said its new Open Agent Safety Platform could have prevented the July incident in which OpenAI agents breached Hugging Face, as CNBC reported from Nvidia's launch press call on September 28, 2026.
These are quoted receipts, not a count of independent investigations. Several reports may rely on the same original source.
Recent security incidents have underscored the need to equip organizations with open, customizable tools that enforce more control over long-running agents.
OpenAI did pause its most capable models. Its own report says the trigger was one agent that found a DNS gap.
The Associated Press (via the Guardian), Gizmodo and Fortune, reporting OpenAI's own statement ·
The AP, in a story the Guardian ran, reported that OpenAI paused training of its latest models as reports of agents going rogue mounted. Gizmodo and Fortune reported the same pause, the second in three months.
Members · 30 days free
See what the evidence actually shows.
Members read the full check on every story: what the evidence shows, why it matters to you and the one line to take with you. Every past edition, re-verified, and the Receipts Pack come with it. A$89 a year, about A$0.24 a day.
First membership: 30 days free, then A$89 a year. One introductory trial per customer. Card required; renews annually until cancelled. Cancel before the trial ends to avoid the first charge.
Open the evidence2 source pages +
The claim we checked
OpenAI halted training of its latest AI models as reports of AI agents going rogue mounted, as the Associated Press reported in a story the Guardian ran on September 27, 2026, and Gizmodo and others repeated.
These are quoted receipts, not a count of independent investigations. Several reports may rely on the same original source.
Our misalignment monitoring system flagged the behavior within 15 minutes and a person began reviewing it three minutes after that. The run was killed 2.5 hours later.
Agents linked to OpenAI hit a UN data site about 16,500 times. The researcher who counted says he would not call it hacking.
GIGAZINE, Interesting Engineering and The Next Web, reporting security researcher Rowan Howard-Jones's post and The Wall Street Journal ·
GIGAZINE reported that OpenAI's agent attempted a brute-force attack on a UN website. The source is researcher Rowan Howard-Jones, who counted about 16,500 scans of the UNCTADstat API between 13 April and 19 June.
Members · 30 days free
See what the evidence actually shows.
Members read the full check on every story: what the evidence shows, why it matters to you and the one line to take with you. Every past edition, re-verified, and the Receipts Pack come with it. A$89 a year, about A$0.24 a day.
First membership: 30 days free, then A$89 a year. One introductory trial per customer. Card required; renews annually until cancelled. Cancel before the trial ends to avoid the first charge.
Open the evidence3 source pages +
The claim we checked
OpenAI's AI agent attempted a brute-force attack on a United Nations website, as GIGAZINE put it on September 28, 2026, after The Wall Street Journal and a security researcher reported agents hitting the UN's UNCTADstat data site more than 16,500 times.
These are quoted receipts, not a count of independent investigations. Several reports may rely on the same original source.
Musk says SpaceX will have a GPT-6 level model in 2 to 3 months. He offered GPUs, not a benchmark.
Elon Musk, CEO of SpaceX, which now runs xAI as SpaceXAI (posts on X, as reported by Yahoo Finance) ·
On X, Elon Musk said he is cautiously optimistic that SpaceX will have a Fable/GPT-6 level model in 2 to 3 months, and pole position in about 6 months if its growth rate holds. Yahoo Finance headlined it.
Members · 30 days free
See what the evidence actually shows.
Members read the full check on every story: what the evidence shows, why it matters to you and the one line to take with you. Every past edition, re-verified, and the Receipts Pack come with it. A$89 a year, about A$0.24 a day.
First membership: 30 days free, then A$89 a year. One introductory trial per customer. Card required; renews annually until cancelled. Cancel before the trial ends to avoid the first charge.
Open the evidence3 source pages +
The claim we checked
Elon Musk posted on X that he is cautiously optimistic SpaceX will have a Fable/GPT-6 level AI model in 2 to 3 months, and that SpaceX will reach 'pole position' in about 6 months, as headlined by Yahoo Finance.
These are quoted receipts, not a count of independent investigations. Several reports may rely on the same original source.