AI Uniti

Europe has started building doctrine for attacks that never cross a line.

Last week Ursula von der Leyen proposed an EU Emergency Security Protocol for incidents that sit below the threshold of armed conflict, and a European Security Council to sit above it. The logic is that the dangerous category is no longer invasion. It is the deniable, coordinated, hard-to-attribute action that does damage without ever triggering a formal response.

Corporates face the commercial version of exactly that.

Nobody breaches your perimeter. Nobody sends a demand. A few thousand accounts behave in ways real users do not, a fabricated clip gets a credible-looking home, and three days later your share price, your tender or your licence to operate is the thing being negotiated.

It sits below every threshold you have. It is not a cyber incident, so security does not own it. It is not a media enquiry, so comms sees it late. It is not litigation, so legal is not looped in. By the time it is any of those things, it is already the accepted version of events.

Governments are writing playbooks for this category. Most boards have not yet named it.

Start with three questions: who owns manufactured narrative risk here, what would trigger a response, and how fast would we know?

Coordination is visible 6 to 12 hours before a story breaks. That window is only useful if someone is accountable for it.

Book a 15-minute Signal by AI Uniti demo: aiuniti.com/signal

#NarrativeRisk #HybridThreats #RiskManagement #Governance #ThreatIntelligence

3 days ago | [YT] | 6

AI Uniti

CISA and the FBI: attackers time attacks to weekends and holidays, when response staffing is thinnest.

Narrative attacks work the same way. A Friday 6pm post has until Monday.

Book a 15-minute Signal by AI Uniti demo: aiuniti.com/signal

4 days ago | [YT] | 2

AI Uniti

War moves fast. Fakes move faster.

As the Strait of Hormuz escalates again, expect a wave of AI-generated footage. In the last round, one fake clip took 7 million views and a fabricated video of a burning US carrier spread widely before it was debunked.

Signal by AI Uniti detects coordination 6 to 12 hours before conventional monitoring.

Book a 15-minute Signal by AI Uniti demo: aiuniti.com/signal

#NarrativeRisk #Deepfakes #Disinformation #RiskManagement

6 days ago | [YT] | 9

AI Uniti

The Fed just raised rates for the first time since 2023.

The same week, Russia's Matryoshka network was caught flooding X, Bluesky and TikTok with celebrity deepfakes and fake CNN reports.

That is not two stories. It is one risk.

Rate shocks create fear. Fear creates attention. Attention is exactly what coordinated campaigns are built to hijack.

The playbook being run on voters works just as well on shareholders, depositors and customers: a fabricated clip, a hashtag, a few thousand inauthentic accounts. The narrative is moving before your comms team has seen it.

We have analysed 793,000 coordinated campaign videos. The pattern holds: coordination is visible 6 to 12 hours before conventional monitoring flags it.

For CROs, CCOs and General Counsel, those hours decide whether you respond or run a post-mortem.

Book a 15-minute Signal by AI Uniti demo: aiuniti.com/signal

#NarrativeRisk #Deepfakes #BehaviouralIntelligence #RiskManagement #Disinformation

1 week ago | [YT] | 13

AI Uniti

Security has CVSS. Every vulnerability gets a number, and the number decides who gets woken up.

Communications has nothing. A rumour arrives, and the response is decided by whoever is loudest in the room, or by how personally annoyed the chief executive is. Two identical incidents get opposite treatment in the same organisation in the same quarter.

There is a scale. Ben Nimmo published it through Brookings in 2020 and called it the Breakout Scale. The platforms that actually run takedowns use it. Almost nobody in a corporate risk function has heard of it.

Six categories, and it measures one thing only: how far the thing has travelled.

One. It exists on a single platform and never leaves the community it was planted in.

Two. It either spreads beyond that community but stays on one platform, or it appears on several platforms and spreads on none of them.

Three. It breaks out on multiple platforms, but mainstream media has not touched it.

Four. It leaves social media entirely and gets reported by the mainstream press.

Five. Celebrities or public figures amplify it, particularly if they endorse it.

Six. It triggers a policy response, or concrete action, or it includes a call for violence.

Now the part worth taking to your next risk meeting.
The overwhelming majority of what crosses your desk is Category One or Two. It is in one place, it is going nowhere, and the people who have seen it already believe it or already do not.

At that level, the highest-risk action available to you is responding. A corporate denial is a cross-medium breakout event that you performed on yourself. You have taken something at Category One and handed it Category Four, complete with your own quote and your own logo, to an audience that had never encountered it.

I have watched organisations do this to themselves and then describe the resulting coverage as an attack.

The scale also tells you where the money actually is. Everything below Category Four is noise you should be measuring and not touching. Category Four is where analysts start asking questions, Five is where it reaches your customers, and Six is where a regulator, a minister or a court gets involved and the costs stop being reputational.

So the only three questions that matter when something lands are:

Where is it on the scale right now?

What is its velocity, and is it moving up or sitting still?

Is it being pushed, or is it travelling on its own?

That last one is the one nobody can answer, which is why the first two get answered badly. If it is being pushed by a small number of accounts that arrived together, it will collapse without you, and responding is the only thing that can save it. If real people are carrying it, no takedown will help and you have a substantive problem, not a manipulation problem.

Different problems. Opposite responses. Identical appearance on the page.
Book a 15-minute Signal by AI Uniti demo: aiuniti.com/signal

Source: Ben Nimmo, "The Breakout Scale: Measuring the impact of influence operations", Brookings, September 2020.
#RiskAndCompliance #CrisisCommunications #NarrativeThreatIntelligence

1 week ago | [YT] | 11

AI Uniti

49.9 per cent of new online articles are now primarily AI-written. In 2020 it was 5 per cent.

Super-recognisers, the best face-readers alive, catch 41 per cent of AI faces untrained. They also call real faces fake a third of the time.

The artefact is not the evidence. The pattern across accounts is.
aiuniti.com/signal

1 week ago | [YT] | 10

AI Uniti

An IDC research vice president made a point about our category that I have not been able to stop thinking about.

The business case for narrative threat intelligence is always written about the company. Reputation, market capitalisation, customer trust. That is what gets it into a budget.

He pointed at the part nobody writes down. When an organisation deploys this at enterprise level, its executives quietly get something else: their own names, faces and voices are now monitored for the same coordinated targeting and deepfake impersonation aimed at the brand. For anyone who has watched a colleague's career damaged by a synthetic video, that is not an abstract capability. In his words, it is “personal risk reduction”.

IDC has published an independent Market Note examining AI Uniti and the emergence of this category. We are the company named in the title. We did not write it and we do not control it, which is precisely why it is worth reading.

Three numbers from inside it.

IDC surveyed 500 organisations across Asia Pacific. 56.2% now rank AI-generated disinformation campaigns, aimed at reputational or operational disruption, among their top AI-driven threats. That sits alongside AI-enhanced phishing and impersonation at 58.4%. Disinformation has stopped being a communications problem and become a security one.

40.6% say explainability, hallucinations and false positives are what stops them adopting AI security tooling at all. That is a market asking for evidence rather than confidence scores.

And by 2028, IDC predicts 85% of organisations in the region will face phishing that uses synthetic identities blending real and AI-generated information.

The note is free. No form, no email address, no gate. Just the PDF.
aiuniti.com/analyst/idc-market-note

Authors: Sakshi Grover, Senior Research Manager, Cybersecurity Products and Services, IDC Asia/Pacific, and Craig Robinson, Research Vice President, Security Services, IDC.

#NarrativeThreatIntelligence #Deepfakes #ThreatIntelligence

2 weeks ago | [YT] | 8

AI Uniti

The most effective disinformation strategy of the last century was written down in a memo, and it is only four words long.

In 1969, an executive at the tobacco company Brown & Williamson set out the plan in an internal document titled Smoking and Health Proposal. Doubt is our product, it says, because doubt is the best way to compete with the body of fact sitting in the public's mind.

Read the rest and it gets sharper. The memo concedes that inside the industry there is no controversy at all. The objective is not to convince anyone that smoking is safe. It states plainly that they cannot take that position. The objective is narrower and far more achievable: establish a controversy at the public level.

Make the question look open. That is the whole strategy. It bought roughly forty years.
Now think about what that means for how you monitor your own brand.
A manufactured controversy does not register as an attack. It does not read as negative sentiment, because it is not negative. It reads as balance. Two sides. Legitimate debate. Reasonable people disagreeing. Every tool built to flag hostility will score it as healthy discussion, and every dashboard will look fine.
That is not a gap in the tooling. It is the point of the tactic.

What has changed since 1969 is only the cost. That campaign needed advertising budgets, friendly journalists and a research committee. Manufacturing the appearance of disagreement now needs none of those things, and the accounts doing it will look like ordinary people with ordinary opinions.
So the question is not whether people are saying negative things about you. It is whether someone is currently manufacturing the impression that the question about you is still open.

Would your monitoring catch that, or would it just look like a balanced conversation?
Source: Brown & Williamson, Smoking and Health Proposal, 1969, in the public tobacco document archives.

#NarrativeThreatIntelligence #ReputationRisk #BrandRisk

2 weeks ago | [YT] | 9

AI Uniti

In February 2025 a tailings dam failed at a Chinese-operated copper mine in Zambia. Toxic waste reached local water supplies. A genuine and serious environmental incident.

The coordinated campaign that followed did not run in Zambia. It ran in Kenya.
More than 30 bot accounts activated inside Kenyan information spaces, amplifying the Zambian incident and framing it as a reason for Kenyan businesses and government agencies to reject partnerships with Chinese mining companies. Kenya had no involvement in the incident. No Kenyan communities were affected. There was no local grievance to draw on, and none was needed.

That is the part worth sitting with if you run a resources business. A campaign against your project does not require anything to have happened at your project.
Two more findings from the same research, which covers coordinated activity across the DRC, Niger, Mali, Rwanda, Guinea and Côte d'Ivoire.

Bot activity around the Loulo-Gounkoto gold operation in Mali surged by more than 400% around trigger events, including the restart announcement earlier this year, and again after Mali revoked foreign mining permits in October 2025.

And those surges are not improvised. The infrastructure sits pre-positioned and activates on anticipated events: regulatory decisions, permit rulings, restarts, price movements. By the time manual monitoring identifies a campaign, it has usually already reached the audience it wanted.

Which gives resources companies something practical to do this morning. Open the corporate calendar for the next twelve months. Permit decisions, environmental submissions, restarts, quarterly results, community consultations.

That is not only your calendar. It is the campaign calendar, and somebody else already has a copy.

#NarrativeThreatIntelligence #Mining #Resources

2 weeks ago | [YT] | 7

AI Uniti

A think tank listed three Council on Foreign Relations experts as affiliates. None of them had ever heard of it.

The International Burke Institute presents itself as an Israel-based expert community. Research papers, a proprietary sovereignty index, a roster of named affiliates. Much of the academic work on the site appears to be copied and misattributed. The site claims UNICEF as a partner institution, which UNICEF denies. No founder is named. It is still live and still posting.

CFR published the analysis yesterday. IBI sits at the centre of a Russian influence operation OpenAI exposed last week: operators logging into ChatGPT over VPNs, generating English-language posts that surfaced on X, LinkedIn, Facebook, Substack and Telegram, and instructing the model to strip any linguistic tells that they were Russian. The posts promoted IBI and its index heavily. The operators also tried to seed fabricated papers by fictitious authors onto genuine academic hosting platforms.

Now the part that should interest every communications and legal team.
The reach was negligible. OpenAI rated it at the low end of level three on a six-level scale. Almost nobody saw the posts.

It still worked, because engagement was never the objective. Citability was. Build something that looks like an institution, furnish it with borrowed credentials, and wait for one journalist, one submission or one policy paper to cite it a single time.
Your monitoring will not catch this. Volume thresholds never trip. Sentiment never turns, because your brand is not being attacked. It is being endorsed by an organisation you have never heard of.

Most organisations watch for who is criticising them. Very few watch for who is claiming them.

Analysis: Jessica Brandt, Council on Foreign Relations, 3 September 2026.
#NarrativeThreatIntelligence #ReputationRisk #GeneralCounsel

2 weeks ago | [YT] | 11