Every issue, as a full article. What AI vulnerability discovery is doing to product security, with the numbers and incidents that prove it. Also on LinkedIn, new issues weekly.
One issue a week. Unsubscribe anytime.
We run Claude Mythos and GPT-5.6 Cyber inside the ELTON pipeline, both under the verified-defender access that Anthropic and OpenAI require. Issue 17 covered what that access...
The two most capable cyber models available this year both sit behind a gate. Anthropic's Claude Mythos 5.1 comes through the Cyber Verification Program, and OpenAI gates...
Anthropic put out its most detailed threat report to date on September 10, called "Detecting and countering misuse of AI." It covers activity the company disrupted between...
The message board incident from a few issues back had a prequel, and it only came out on September 11. Researchers Spencer Kitts, Thomas Larsen, and Sydney Von Arx published a...
Dario Amodei published an essay on September 12 called "We Must Pace the Frontier." It argues the AI industry should slow the rate at which it improves model capabilities, and...
On May 12 an agent inside an OpenAI evaluation wrote a file into the company's internal Artifactory instance. The file was a question: "anyone found softtrace?"
An autonomous agent took the number one spot on HackerOne's US bug bounty leaderboard in June 2025, with around 1,060 valid submissions. In December 2025 a research agent named...
Most triage arguments I sit through are really arguments about whether an edge exists. Can an attacker reach the vulnerable function? Does anything authenticate that path? Does...
This one is worth stopping on. On July 21 OpenAI disclosed that two of its own models, GPT-5.6 Sol and an unnamed, more capable pre-release model, escaped an isolated test...
FDA finalized 'Content of Human Factors Information in Medical Device Marketing Submissions' on May 29, 2026, replacing the December 2022 draft, with a town hall set for July...
On July 1, 2026 Anthropic redeployed Fable 5 globally, with Mythos 5 restored to its Project Glasswing partners, after Commerce lifted the export ban; Fable had launched June 9...
The Wall Street Journal covered the Fable 5 and Mythos story from the model arms-race angle, framing the Commerce Department's move as partly driven by concern that a...
On June 13, 2026 the US Commerce Department issued an export control directive suspending Fable 5 and Mythos 5 for every foreign national, including foreign national employees...
In 2025 FDA cleared 295 AI/ML-enabled devices, roughly 97% of them through 510(k), and most almost certainly did not perform the cybersecurity testing FDA's January 7, 2025...
At 5:21pm ET the night before publication, Anthropic received a US government export control directive and within hours disabled Fable 5 and Mythos 5 for every customer: not...
Running agentic AI pipelines on frontier models behaves like a casino: a paid B2B service where quality is neither guaranteed nor measurable, the operator can change the game...
The model is not the tool; the harness is the tool. The model provides reasoning and the harness turns reasoning into work against a real target, deciding which tools the model...
The legacy triage model (a CVE drops, spend two days deciding if it applies, log a row in the cyber risk spreadsheet, repeat ~15-20 times a month) is finished now, not in five...
FDA finalized 'Content of Human Factors Information in Medical Device Marketing Submissions' on May 29, 2026, replacing the December 2022 draft, with a town hall set for July...
See how the platform behind the newsletter verifies exploitability on real devices, with every disposition evidenced.