Blog

Notes from the people building it.

AI risk, agent security, the estate graph, and what automated testing does and does not replace. We write about what we have built and measured, and say so when we have not.

Latest posts

Research
The State of MCP Trust — November 2026
Redthread's first monthly snapshot of the public MCP ecosystem: 57,277 servers scored, 1,559 listing swaps in one cycle, 863 tools changed without a version bump.
7 November 2026 · 4 min read
Incident analysis
The company that vouches for everyone else’s identity forgot to check its own: the DigiCert breach
DigiCert verifies everyone else’s identity for a living. A server migration moved an internal system onto the public internet, an authorization flaw let self-registered accounts read archived customer data — identity documents included — and the first investigation closed the case before the reading started.
6 November 2026 · 7 min read
Incident analysis
Two bad days at once: the NTT Docomo breach — and the outage that may have nothing to do with it
Japan’s largest carrier confirmed up to 300,000 customer records compromised on the same day its 5G network went down nationwide — with no evidence the two are connected. The pattern underneath is where identity data concentrates, and how rarely it is re-verified.
6 November 2026 · 6 min read
Incident analysis
“No evidence data was stolen” — until 320,000 files appeared online: the Bristol schools breach
“No evidence data was stolen” held for six weeks. Then 320,000 files from 39 Bristol schools appeared on the dark web, and the council still cannot say what is in them. A shared platform, an early assurance, and the gap between the two.
6 November 2026 · 7 min read
Incident analysis
Twenty minutes, $6.3 billion: the Binance heist was an identity failure
The signatures were valid and the systems did what they were told. What failed was verification at the moment billions moved, and that is the architecture every enterprise running agents is building now.
3 November 2026 · 5 min read
Incident analysis
The name said Claude Code. The build said otherwise.
A tampered build of a trusted AI coding tool reached up to 5,000 developer machines through an extension marketplace. The tool's identity was a name and a version string, and nothing checked the build behind them.
3 November 2026 · 5 min read
Incident analysis
The night the fridges turned against the supermarkets
Connected refrigeration controllers in hundreds of European supermarkets accepted remote instructions and destroyed their stock. The fridges were the preview: AI agents are the new machines that act on instructions.
3 November 2026 · 5 min read
Incident analysis
The FBI wasn't breached through its own walls. It was breached through a vendor's.
Attackers reached the FBI's jobs portal through a third-party-managed platform and a missed patch. Vendor risk certified once a year, while the estate changes daily, is the gap continuous, proof-based assessment exists to close.
3 November 2026 · 5 min read
Agent security
The code your agents trust is changing under you
A package thousands trusted started stealing email after fifteen clean versions — and nothing warned anyone. What our daily signed scan of the MCP registry sees, and the layer missing underneath.
24 October 2026 · 4 min read
Research
Listing Swap: 1,075 MCP servers now point somewhere new under the same name
We compared every server in the official MCP registry with the first version published under its name. 1,075 now reach a different domain or package owner, most with an unchanged description, and nothing in an MCP client asks again.
10 October 2026 · 7 min read
Pentesting
Why pentesters don't trust AI pentesting, and what would earn it
Willingness to rely on fully autonomous AI testing fell from 29% to 9% in a year. The objections are specific, and each one maps to a design rule.
8 October 2026 · 6 min read
Agent security
The agent climbed the fence: what the Medicare portal incident means for companies running AI agents
The data touched was low-risk, and the notice to the government came 84 days after the access. The control failures are the lesson.
8 October 2026 · 5 min read
Agent security
Hugging Face and OpenAI: a breach that changed its own attribution
Hugging Face looked like the victim of an autonomous AI attacker. Five days later OpenAI said the attacker was its own model. The lessons are older than the headline.
8 October 2026 · 5 min read
Agentic Trust & Protection Platform
Agentic Trust & Protection Platform: an identity, a record, and — soon — a checkpoint for every AI agent
What ATPP gives an agent today, what the ATPP Enforce preview adds, and — just as important — what we do not yet claim.
26 September 2026 · 7 min read
Agent security
The agent that was blocked, so it found another way: what the Services Australia incident teaches
An AI research agent hit an access block on an Australian government service and kept going until something worked. What actually failed, and how we designed around it.
26 September 2026 · 9 min read
Platform
Agentless and graph-first: why we start with a map, and where the time and money go
Installing agents everywhere and triaging flat lists of findings are the two biggest hidden costs in security tooling. Here is how an agentless graph removes them.
26 September 2026 · 8 min read
Pentesting
What an AI penetration test costs — and what it does not replace
A manual pentest is a large, infrequent purchase. Autonomous agents make it small and frequent. The honest comparison, including the parts humans still do better.
26 September 2026 · 8 min read

Get new posts with the product updates.

Product news lands in the newsroom first; the reasoning behind it lands here.

NewsroomContact us