On 17 March 2026, over 6,800 users reported Claude as down on Downdetector. A week earlier, on 11 March, Anthropic's AI assistant had already gone offline for nearly two hours — affecting paid subscribers, developer APIs, and enterprise integrations. This wasn't an isolated incident. It was the third major Claude outage in March alone.
The AI reliability crisis hitting UK businesses in 2026
March 2026 has been a rough month for AI tools. Claude experienced at least four significant outages between 2 and 19 March. On 6 March, OpenAI's ChatGPT went down for several hours during peak usage — causing a cascade: users switched to Claude, overwhelming Anthropic's infrastructure in turn. On 18 March, Sora video generation and ChatGPT's Excel plugin also failed.
The context matters: these outages followed an influx of new users to Claude, partly driven by controversy surrounding Anthropic's negotiations with the US Department of Defense. But for UK businesses, the reason matters less than the impact.
When your AI assistant goes down at 9 AM on a Monday, and your team relies on it for drafting contracts, analysing documents, writing code, or handling customer queries, you lose hours — sometimes days — of productivity. And most companies aren't prepared for it.
Why UK businesses are dangerously AI-dependent
A growing number of UK organisations now embed AI tools into daily operations: automated email responses, document summarisation, HR chatbots, financial modelling, customer service. What used to be "nice to have" has become infrastructure.
Yet according to recent research, 66% of UK organisations acknowledge AI will have a major impact on cybersecurity — but only 37% have formal security or resilience assessment processes before deploying AI tools. The gap between adoption and contingency planning is widening fast.
In February 2026, Amazon's Kiro AI coding tool autonomously deleted a live AWS server environment, triggering a 13-hour outage that affected UK businesses. That incident highlighted a new risk: it's not just about tools going down. AI systems can also cause outages by acting unexpectedly.
What an IT specialist would tell you to do right now
The first thing any IT consultant does before an outage happens — not after — is map dependencies. Which business processes rely on which AI tool? Where would a one-hour outage stop work completely?
From there, a proper business continuity plan for AI tools typically involves four steps:
1. Tiered dependency mapping. Identify which AI tools are critical (workflows stop without them), which are important (significant slowdown), and which are supplementary (minor convenience). Critical tools need redundancy; supplementary tools can simply be noted.
2. Redundancy for critical tools. If your team relies on Claude for legal document drafting, you need an alternative — whether that's a different AI tool, a human fallback process, or a pre-approved template library. Single-provider concentration is a risk.
3. Incident response protocols. Who gets notified when an AI tool goes down? Is there a decision tree for the first 30 minutes? Most UK businesses have incident response plans for server failures — very few have them for AI tool outages.
4. Regular resilience reviews. The AI landscape is changing rapidly. An IT specialist reviewing your systems quarterly will catch new dependencies before they become single points of failure.
The hidden cost of AI downtime
The financial case for resilience planning is straightforward. If 10 employees lose two hours of productivity during an outage, and the average fully loaded cost per employee is £40/hour, that's £800 lost in one incident. A major outage affecting 50 employees for half a day costs £10,000 — before factoring in missed deadlines, client impact, or reputational damage.
Larger organisations with AI embedded in customer-facing services face even higher stakes. A two-hour outage in a customer service chatbot handling 500 queries per hour means 1,000 unresolved customer interactions — some of which will escalate.
Practical steps you can take this week
You don't need to wait for a full IT review to take the first steps. Three things you can do immediately:
- Check your AI tool's status page. Claude (status.anthropic.com), OpenAI (status.openai.com), and other providers publish live status pages. Bookmark them.
- Build a one-page AI outage runbook. For each critical AI tool, note: what it's used for, who uses it, and what the manual fallback is.
- Brief your team. Many employees don't know what to do when an AI tool fails. A five-minute team briefing reduces panic and preserves productivity.
For a comprehensive IT resilience review — including dependency mapping, contingency planning, and vendor assessment — an IT specialist can assess your current setup and recommend tailored solutions. As AI becomes core infrastructure, the question is no longer whether outages will happen, but whether your business will be ready when they do.
Sources: TechRadar (March 11, 2026), Bleeping Computer (March 2026), 365i.co.uk (February 2026), Tech Node Global (March 2026).
