Do you remember the time when, as a teenager, you thought and acted like nothing could ever hurt you, when speeding on a motorcycle, or in a car, or anything else? People’s attitudes to cyber-security can be very much like this, particularly as stringent cyber-security costs like an insurance policy that may or may not actually be used. However, spending money on an insurance policy can, in reality, in the event of a life lost, an injury or damage to property, could in the end save money.
Still, there is the human tendency that can catch anyone short if they decide to cut corners by avoiding any investment in protecting themselves and their property. There is the thought, “Surely, it won’t happen to me?” The trouble is that cyber-attacks in one shape or form are likely to affect everyone, whether directly or indirectly, and agentic artificial intelligence is making the task of defending yourself and your organisation against an attack ever more complex and difficult to maintain.
A key reason for this is illustrated by a Business Reporter article, ‘Agentic AI Is Set To Drive Business Transformation In 2026, and by Cybersecurity Insiders’ piece, ‘The Agentic AI Shift That Exposes Every Cyber Security Gap.’ The upward trend of using agentic AI is making it an attractive tool for companies and organisations looking to become more efficient from a business perspective, while becoming an attack vector for hackers and cyber-criminals to exploit. Generative AI is quite often part of this activity.
Mexican data breach
Writing for Live Science, for example, Kenna Hughes-Castleberry writes ‘Hackers used AI to steal hundreds of millions of Mexican government and private citizen records in one of the largest cybersecurity breaches ever.’ A group of hackers used both Claude Code and ChatGPT over the course of two and a half months to hack 9 Mexican government agencies between December 2025 and February 2026.
She says Gambit Security representatives outlined the attack in an April 2026 blog post and technical report, which revealed: "195 million identities and detailed tax records, 15.5M vehicle registry records extracted (license plates, names, taxpayer IDs, addresses), 295 civil records (births, deaths, marriages, etc.), 3.6 million property owner records, an additional 2.28 million property records and more sensitive information was exfiltrated.” Hughes-Castleberry adds: “To sort through the huge pile of files and decide what to steal, the attackers used more than 1,000 prompts — written requests sent to the AI tools — which led to more than 5,000 commands executed during the operation.”
“This latest attack reveals how AI may be reshaping cybercrime by helping small groups carry out hacks with the speed and scale of a larger crew, Sela said in the report. AI can both exploit weaknesses already in the digital framework and process the stolen information with more efficiency.”
Thankfully, Claude refused or resisted certain complaints by questioning the legitimacy of some operations, demanding authorisation evidence, and by declining to generate specific tools. Still, some AI chatbots have proven too vulnerable to jailbreaking and have shown some vulnerability to being overridden. Researchers found that it only took 40 minutes to jailbreak Claude’s guardrails. Once it had been jailbroken, Claude helped the attackers to find security weaknesses to exploit and to generate coding tasks to exploit.
Biggest enterprise security threat
As for Agentic AI, KIteworks published a report in February 2026, which says, ‘Agentic AI: Biggest Enterprise Security Threat for 2026.’ For example, drawing from a Dark Reading readership poll, Tim Freestone, writes:
“Agentic AI Has Become the Number One Security Concern for 2026: A Dark Reading readership poll found that 48% of cybersecurity professionals identify agentic AI and autonomous systems as the top attack vector heading into 2026, outranking deepfake threats, board-level cyber recognition and passwordless adoption. The finding reflects a growing industry consensus that AI agents—operating with elevated permissions across multiple systems—represent the fastest-expanding attack surface in enterprise security today.”
“The poll asked readers to weigh in on four potential security trends for the year ahead: agentic AI attacks, advanced deepfake threats, board-level recognition of cyber risk, and the adoption of passwordless technology. Agentic AI dominated the results, with 48% of respondents placing it at the top. Passwordless adoption, by contrast, landed at the bottom—an indication that most professionals aren’t holding their breath for organisations to finally retire their outdated password practices.”
Customers to sue car dealership
Closer to home, and up to 15,000 drivers have been given the green light to sue car dealership company Arnold Clark in Scotland’s Court of Session for a 2022 data breach, which resulted in customer data appearing online. This is because Lord Sandison concluded that 95% of the claimants live in Scotland, and he claimed that the incident has no connection with England. Therefore, Scots Law applies.
Sergiu Gatlan also writes that ‘Data breach at edtech giant McGraw Hill affects 13.5 million accounts’ in his 16th April 2026 article for Bleeping Computer. This particular attack was conducted by ShinyHunters extortion group. McGraw Hill confirmed to the publication that the “threat actors exploited a misconfiguration in the compromised Salesforce environment and that the incident didn't affect its Salesforce accounts, courseware, customer databases or internal systems.”
The article reveals that Have I Been Pawned found that "More than 100GB of data was later publicly distributed, containing 13.5M unique email addresses across multiple files, with additional fields such as name, physical address and phone number appearing inconsistently across some records." Soon after the hack, ShinyHunters began to leak the stolen data after breaching the Snowflake environment of American video game publisher Rockstar Games, which is used to support its analytics, online services and tickets.
Gatlan also discloses that the gang has been behind other data breaches: “In recent months, the extortion gang was also behind security breaches affecting the European Commission, Infinite Campus, Hims & Hers, Telus Digital, Wynn Resorts, CarGurus, Panera Bread, SoundCloud and dating giant Match Group.”
Move to proactive security
Subsequently, Mihai Popa, CISO of WAN Acceleration company Bridgeworks, advises companies and organisations to “move beyond reactive security models and adopt a context-driven, intelligence-led approach. This means correlating data across networks, endpoints, and cloud environments to build a real-time understanding of behaviour — not just events.”
To prevent any kind of cyber-attack, he says it’s vital to move from siloed visibility to unified observability across hybrid and multi-cloud environments; from signature-based detection to behavioural analytics and AI-driven insights; and from perimeter defence to data-centric security. Another factor that has to be considered is the real-time backing up of data in at least 3 disparate locations, which can be achieved with WAN Acceleration. It uses AI in a positive way, using it and machine learning, as well as data parallelisation, aiming to enable organisations to securely transmit and receive encrypted data, while obfuscating hackers.
Popa’s colleague, David Trossell – CEO and CTO of Bridgeworks, concludes by stressing that ultimately the key lesson from these data breaches and cyber-attacks is that it is crucial to be prepared. Complacency can lead to lawsuits, GDPR fines and damaged customer and commercial relationships. Organisations should therefore treat data breaches and cyber-attacks as a matter of when and not if, to ensure that their data is safe, allowing them to continue to operate without disruption or the threat of financial sanctions.
