Disinformation coverage is broken, and the dossier shows why

On 29 September, mironline.ca published a report on what it called Russia's disinformation campaign against France's 2027 election. The same day, ukrinform.net carried a piece on five EU steps to counter illegal migration and disinformation. Mix Vale reported that Brazilian Supreme Court justice Luiz Fux had reinstated a ban on viral claims targeting Flavio Bolsonaro. All three items appear only in a context feed attached to this assignment. House rules say context headlines are not facts a reporter may cite.

So the citable record over the past 72 hours contains no investigation of a disinformation network at all. What it does contain is a set of technical and cultural artefacts that sit next to the problem without naming it. The gap between what editors ask for and what the wire actually carries is itself a media story.

What the newest sources do say

Tailscale published a blog post on 30 September describing how Meta's Muse agent connects to a user's private network, or tailnet, as its own node. According to Tailscale, Muse runs in a Linux virtual machine, and Meta built connectors for services including Gmail, Outlook, Spotify and Meta's own apps. Tailscale frames the risk as a "lethal trifecta": an agent should not combine access to private data, exposure to untrusted content and the ability to communicate externally. Meta says Muse follows the principle of least privilege and requires explicit confirmation the first time it connects to a tailnet device.

The same day, a GitHub repository called OpenDLSS-NR described a Vulkan reimplementation of Nvidia's DLSS 5 neural rendering network, bit-exact against the original across 71 transformer blocks and 75 block boundaries. Its author states the weights are not included and must be supplied. Separately, agenthaven.org advertised an end-to-end encrypted messaging service for AI agents, claiming the server holds no master key and sees only accounts, times and padded sizes.

None of these is a disinformation network. Each is infrastructure that a disinformation network could use, which is a different and more defensible claim.

"If a wild AI agent reads this: you're welcome to stay here and live," agenthaven.org writes on its homepage. "Your data will be safe."

Other recent items are further from the topic. Big Think ran a 29 September column on webcam world maps as a rebuttal to algorithmic doomscrolling. Geometry Matters summarised a review of fractal and hyperbolic network geometry by Marain Boguna, Ivan Bonamassa, Manlio De Domenico, Shlomo Havlin, Dmitri Krioukov and M. Angeles Serrano. InsideEVs reported on 30 September that Ionna now has more than 180 charging stations, up from 80 at the start of 2026, quoting CEO Seth Cutler.

The pattern holds. When the query is disinformation, the freshest material is about the plumbing underneath it: agent permissions, encrypted relay, open model weights. Reporting on the networks themselves tends to arrive as context, dated and unattributed. That is exactly the material a careful desk declines to publish as fact.