As AI becomes woven into critical business functions, data governance, protecting the integrity of AI models, third-party risk and regulation are becoming very difficult to manage. Thatโs according to a report by AXA XL, the property and casualty and specialty risk division of the insurance company AXA, and the intelligence and corporate cyber security consultancy S-RM. In their studyย Building Resilient AI: Managing AI risk through governance, security and resilience, they warn that AI is becoming embedded in critical business processes faster than many can adapt their governance, security and incident response. They call on business to treat AI risk as an enterprise resilience issue rather than solely a technology, or compliance, concern.
The report describes AI as โa double edged swordโ, creating more potential entry points for cyber threats and introducing new forms of operational, regulatory and third-party risk.
Among other things, the report suggests businesses set accountability for AI across the enterprise, including formal deployments, embedded software features and shadow AI. Businesses ought to protect sensitive data and strengthen identity and access controls, as AI systems become more autonomous; and manage AI risk throughout its lifecycle, from data collection and model development to deployment, monitoring and incident response. Also called for are due diligence on AI vendors and preparing for AI-related loss scenarios whether in terms of cyber, fraud, liability or business interruption.
Jonathan Salter, Head of Risk Consulting at AXA XL, said: โAI is moving from experimentation into the systems and processes organizations rely on every day, but governance is not always keeping pace. The organizations best placed to capture AIโs value will be those that know where it is being used, understand the business consequences when it fails and build security and resilience into deployment from the outset.โ
And Rebiah Bardot-Girard, Head of Cyber Risk Consulting Services at AXA XL said: โAI risk rarely emerges in isolation. It amplifies existing weaknesses in identity management, data governance, supplier oversight and incident readiness. A practical inventory of where AI is used, what data it can access and where it can take or influence action is now a fundamental starting point for resilience.โ
The report covers risks including data leakage, model manipulation, prompt injection, unreliable outputs, shadow AI and overly autonomous agents. It points to the World Economic Forumโs findings, that 64 per cent now assess the security of AI tools before deployment, up from 37pc a year earlier. As AI gains access to sensitive data, business applications and decision-making processes, businesses remain exposed unless oversight continues throughout deployment and operation.
The report,ย Building Resilient AI: Managing AI risk through governance, security and resilience, is available at the AXA XL website: https://axaxl.com/. AXA is acquiring S-RM.
Retail findings
Netskope Threat Labs has found that regulated data, including sensitive information such as personal or payment data, accounts for 56 per cent of AI-related data policy violations in the retail sector. Source code accounts for a further 20pc of violations, while passwords and API keys make up 16pc. This comes as AI becomes more deeply embedded across retail. The proportion of employees actively using AI applications has risen from 39pc to 65pc over the past year, while near all, 97pc now use applications containing AI-powered features and 90pc interact with AI systems that use customer or user data to train models.
Retailers are responding by trying to bring more of that activity under corporate control, with organisation-managed AI adoption rising from 40% to 73%. Personal AI use has fallen from 70pc the previous year, but remains significant with 44pc of employees using personal tools. The boundary between the two is also becoming less clear, the study suggests, with the proportion of users switching between personal and enterprise accounts rising from 11pc to 18%pc.
The way AI interacts with business systems is also becoming more complex. The number of AI agents interacting with remote MCP servers increased by around 400pc, while MCP-related activity rose by about 300pc. These connections allow AI systems to interact with external tools and data sources, meaning retailers will increasingly need to understand not only which AI applications are being used, but what information those systems can access.
Embedded
Gianpietro Cutolo, Cloud Threat Researcher at Netskope said: โRetailers are moving beyond simply experimenting with AI and are starting to embed it across everyday operations, customer experiences, and business workflows. But as AI becomes more deeply connected to the data and systems that power the retail business, the risks become harder to separate from the opportunities. The challenge is no longer deciding whether to use AI, but making sure it can be used at the speed the business demands without losing control of sensitive customer and company data. Retailers that can combine rapid AI adoption with strong visibility and governance will be best placed to turn AI into a competitive advantage without creating unnecessary risk.โ
Findings include:
-most, 85 per cent of AI-related violations are upstream (involving data sent into AI tools), compared with 9.5pc downstream (involving AI outputs) and 2.5pc related to content filtering.
-Retail users encountered and clicked malicious links returned by AI applications at rates ranging from around 40 to more than 160 per 100,000 users per week.
-Anthropic Claude Platform is now used by near all, 96pc of retail firms in the dataset, ahead of ChatGPT at 84pc.
Small businesses
A 2026 SMB (Small-to-Medium-sized Business) Cyber Risk Report by the cyber firm ESET shows that many in the UK are still exposed to common attack techniques despite growing concerns over more sophisticated, AI-powered threats.ย The cyber firm surveyed 500 small and medium businesses (SMBs) in the UK, and found a gap between security concerns and the everyday risks. Almost half (49pc) of UK SMBs experienced a cyber incident last year. The most common causes were phishing, unpatched vulnerabilities, weak passwords, and a lack of monitoring. On average, it took a UK small business just over four weeks to identify and recover from a breach. The top security concern for those small businesses: AI-powered malware.
Jake Moore, Global Cybersecurity Advisor at ESET said: โItโs true that AI will help cybercriminals to speed up their attacks against UK businesses. But the core tactics used like phishing or exploiting software vulnerabilities will largely stay the same. Until organisations can improve their security by locking down the basics, smaller businesses will be easy pickings for cybercriminals looking to steal sensitive data and turn a profit.โ
Photo by Mark Rowe: from the July 2026 edition of Professional Security Magazine, Jake Moore speaking at the Infosecurity Europe show at London Excel in June.
