AI and cloud briefing: a hacker used AI agents to breach 100+ companies for $5,000, Microsoft and Anthropic turn AI toward cyber defence, and UK fashion retailers face a wave of attacks

Here is our latest round-up of what’s changing in AI and cloud for UK businesses, with links to the original sources so you can read further.

Lead story: a hacker used AI agents to breach more than 100 companies for about $5,000

Reuters has traced a hacking campaign to a single, Chinese-speaking attacker who used AI agents to do most of the work of breaking into company websites. According to researcher Snorre Fagerland at NorForge, who found exposed control panels left behind by the attacker, the setup combined an open-source AI coordination platform called Cairn with what looks like Anthropic’s Claude Code running on a DeepSeek model underneath. The campaign began on 25 September and ran for several weeks before Reuters reported on it on 19 October.

The agents attempted to break into more than 1,200 websites and succeeded at more than 100 companies, mostly e-commerce sites in the US, Canada, Australia and Japan. Reuters named Deere, SharkNinja and Cummins among the affected firms. Fagerland estimates the attacker spent roughly $50 in AI tokens per successful break-in, and about $5,000 across the whole campaign, though he says these figures are uncertain. Deere said it took remediation action and saw no sign that sensitive information was accessed; SharkNinja said a standalone support site was affected and contained; Cummins said it is investigating a possible attack on a website run by a third-party vendor. The developer behind Cairn said on social media that the platform was never intended for hacking. More detail is available from Reuters’ findings as covered by superpowerdaily and Cyber Security News.

Why this matters: this is one of the clearest examples yet of AI agents lowering the cost of hacking at scale, rather than just making individual attacks cleverer. A single person, spending roughly the price of a laptop, reached over 100 companies. If your business runs customer-facing websites or e-commerce systems, this is a good prompt to check that routine vulnerability scanning, patching and web application firewalls are in place, since this kind of automated, low-cost probing is likely to become more common, not less.

Artificial intelligence: Microsoft and Anthropic put AI to work defending, not just attacking

Microsoft has moved its Project Perception system into public preview, a tool that puts AI agents on the defensive side of security operations. The system runs three groups of agents: “red team” agents that look for possible paths an attacker could take, “blue team” agents that investigate what is actually happening and decide which alerts matter, and “green team” agents that make corrective changes to strengthen defences. Microsoft says human operators stay in control throughout, and that businesses choose which actions need human approval and how narrowly each agent is allowed to act.

Separately, Anthropic has launched OSS Scanner, a free service that uses its Claude models to repeatedly scan critical open-source projects for security flaws. The scanner builds each enrolled project in an isolated, offline environment, runs scanning agents against it, has other agents check the findings, and emails maintainers a report with reproduction steps and, where possible, a suggested fix. No human at Anthropic reviews the findings before maintainers see them, and the service is aimed at established, widely used projects that are significant to internet infrastructure. More detail is available from TechRepublic on Project Perception and Cyber Security News on OSS Scanner.

Why this matters: the same AI agent capabilities now being used to attack companies, as in our lead story, are also being turned toward defence, often by the same underlying model providers. For UK businesses, this is a reminder that AI-assisted security tools are becoming mainstream, but also that they still need clear rules about which actions an agent can take on its own and which need a person to sign off.

Cyber security: a wave of attacks hits UK fashion retailers

TFG London, the UK arm of South African retailer The Foschini Group, detected a cyberattack on 11 October that disrupted IT systems supporting its Phase Eight, Hobbs and Whistles brands. Websites and backend order processing were affected, though physical stores stayed open and its Inside Story and White Stuff brands were not hit. The company has brought in outside specialists, reported the incident to UK authorities, and says it has found no confirmed evidence yet that customer or payment data was exposed, with its investigation ongoing.

This follows a breach at ASOS, confirmed on 6 October, where attackers sent unauthorised push notifications through the company’s own app claiming to have stolen customer data from its Snowflake environment. ASOS confirmed that third-party platforms used to message customers were accessed without permission, and that names and contact details may have been exposed, but says it does not believe payment card details or passwords were affected. A group calling itself “Xuanye” claimed responsibility but has provided no evidence. H&M is separately investigating claims by a group called ShinyHunters, though it says it has found no sign that its own systems were compromised. More detail is available from Help Net Security on the TFG London attack and BleepingComputer on the ASOS breach.

Why this matters: three well-known UK or UK-facing retailers have reported cyber incidents within about two weeks of each other. Retail businesses hold large volumes of customer contact data and depend heavily on third-party platforms for things like app notifications and cloud data storage, which is often where these incidents start. If your business uses similar third-party messaging or data platforms, it is worth checking who else has access to them and how quickly you would notice if that access were misused.

The takeaway

Today’s stories show AI cutting both ways in security. The same kind of AI agents that let one hacker reach over 100 companies for a few thousand dollars are now also being offered, by Microsoft and Anthropic, as tools to find and fix weaknesses before attackers do. Meanwhile, the run of attacks on UK fashion retailers is a reminder that plenty of breaches still come down to familiar weak points: third-party platforms, messaging systems and basic access control, not exotic new techniques. Businesses that keep the basics tight, patching, access reviews and monitoring, will be better placed whichever side of this AI-security balance ends up moving faster. We can help you work through what applies to your business. Get in touch.

AI Agent based research and content, AI can do mistake, please help us to improve.