Skip to content

Latest news

Eleutheros Ltd
Business district skyline reflected in a river at blue hour

Digital business trends

AI Blamed for a Wave of IT Outages in the United States

Siècle Digital reports that several US services suffered major outages, with AI singled out as a possible cause while a cyberattack is ruled out in some cases. The report does not specify which systems failed, how long outages lasted, or what role AI actually played.

Siècle Digital reported on 20 November 2026 that several services in the United States were hit by significant outages, and that artificial intelligence is being pointed to as a possible cause. According to the outlet, the cyberattack hypothesis has been set aside in some of these cases. The report also mentions a Microsoft bug that lasted many hours, while stressing that Microsoft was not the only company involved. It does not name the other providers, the services affected, the duration of each incident, or the technical mechanism by which AI is suspected to have played a role.

What the report establishes, and what it leaves open

The confirmed elements are narrow: outages occurred, they affected more than one company, and in some instances an external attack was excluded. Everything beyond that — scope, root cause, customer impact — remains unspecified. That distinction matters for anyone building a risk picture, because "AI is suspected" can describe very different situations: a model generating faulty configuration, an automated remediation system acting on bad signals, or simply a correlation between the timing of incidents and the spread of AI-assisted tooling.

Well-established context helps frame this without overreaching. Over the past two years, engineering teams have wired AI assistants into code generation, infrastructure-as-code, log triage and incident response. Each of these areas touches production directly. A flawed change pushed by an automated agent and a flawed change pushed by a human are equally capable of taking a service down; the difference lies in volume, speed and how quickly the change is reviewed.

Why this lands on marketing and growth teams

Outages are usually treated as an engineering problem, but the commercial exposure sits elsewhere. Checkout pages, sign-up flows, paywall logic, tracking pixels and attribution pipelines all depend on the same infrastructure. When a service degrades, conversion drops silently and campaign spend keeps burning. A Meta or TikTok campaign optimising toward a broken landing page will happily spend the budget and learn the wrong lessons about the audience.

Three checks a team can run this week:

  • Confirm that monitoring covers the revenue path end to end — form submission, payment, confirmation — not just uptime of the marketing site.
  • Verify who is authorised to pause campaigns and budgets during an incident, and make sure that person is reachable outside office hours.
  • Ask engineering which automated agents can change production configuration, and whether those changes require human approval.

The governance question underneath

The uncomfortable part of the story is attribution. When an outage follows an automated action, logs often show the action but not the reasoning behind it, which makes post-incident review slower and blame harder to assign. Until the report's suspicions are substantiated, the safe reading is that AI is a candidate cause among others, not a conclusion. Eleutheros's view is that the practical response is procedural rather than technological: know what is automated, know what it can touch, and know how fast you can stop it.

Also worth reading

All news

Aerial night view of Alexander Nevsky Cathedral in central Sofia

Contact

Tell us where the project stands.

A few lines are enough: what you are building, where it is today, and what is blocking it. We answer every message personally.

Contact us contact@eleutheros.eu