01
A risk category you can actually measure
Coordination is a behavioural fact with an evidence trail, not a judgement about tone. It can carry a likelihood, an impact and a control, and it can be reported the same way quarter after quarter.
Solution
Narrative Attack Detection for Chief Risk Officers
AI Uniti gives risk leaders a way to put coordinated narrative manipulation on the risk register as a measurable category, with deterministic behavioural evidence rather than sentiment scores.
The threat to the CRO
Coordinated narrative attacks sit in the gap in most risk frameworks. They are not a technical control failure, so security does not own them. They are not a market event, so treasury does not own them. They arrive looking like public opinion, so they get filed as a communications problem and leave the register entirely.
That filing decision is the exposure. A manufactured campaign can move a share price, stall an approval, or make a recall look like negligence, and none of those consequences care which department noticed first. The reason narrative risk is rarely on the register is not that risk leaders think it unimportant. It is that a category you cannot measure is a category you cannot report, and sentiment dashboards do not produce evidence a risk committee can act on.
What Signal gives the CRO
Signal monitors X, Bluesky, Mastodon, RSS and YouTube for the behavioural signature of coordination: temporal synchronisation, amplification structure, network relationships and cross-platform activity. It does not ask whether the sentiment is negative. It asks whether the pattern is organic.
That distinction is what makes it reportable. The output is a deterministic evidence chain showing which accounts acted together, when, and how, which is the same standard of evidence a risk function already expects from every other category on the register. Where the exposure is state-aligned or campaign-driven, see foreign interference and coordinated brand attack.
IDC has published a Market Note examining AI Uniti. Read the Market Note.
What you get
02
Deterministic evidence, not a confidence score
Every finding shows the accounts, the timing and the amplification path. When a risk committee asks how the system reached its conclusion, there is an answer that does not begin with "the model".
03
Findings that survive scrutiny, including the negative ones
We have published an investigation that concluded there was no coordination, showing the method and the data that ruled it out. A detection capability that cannot return "nothing here" and show its working is not a control, it is an alarm.
04
Cross-platform by design
Coordination that looks like nothing on one platform is often obvious across four. Single-platform monitoring is how manufactured campaigns are mistaken for organic sentiment.
Frequently Asked Questions
Is this a communications problem rather than a risk one?
Communications owns the response. Whether the event is organic or manufactured is a factual question with evidentiary consequences for regulatory reporting, insurance, litigation and board disclosure, and those are risk questions.
How do we report this to a risk committee?
As a category with a stated likelihood, a stated impact and a named control, supported by the evidence chain behind each finding. The output is built to be quoted in a paper, not screenshotted from a dashboard.
What about false positives?
Deterministic detection means every finding carries the behavioural evidence it rests on, so a finding can be examined and rejected rather than merely trusted. We publish negative results as well as positive ones.
Does this replace anything we already run?
No. It sits upstream of existing monitoring and feeds the workflows you already have.