Security | Threat Detection | Cyberattacks | DevSecOps | Compliance

As regulators tighten expectations and cyber attacks increasingly exploit supply chains, organisations must shift from periodic vendor assessments to continuous third-party cyber resilience. For years, third-party cyber risk management focused primarily on vendor due diligence and annual security assessments. The objective was simple: determine whether a supplier met an acceptable level of security at a specific point in time.

In cyber security, we often say that "we are all in this together". In my role across Channels and Alliances, that principle is a constant. We are stronger and more resilient when we work as one.

For most business technology, the cost is relatively easy to understand. You buy a licence, agree a contract or pay for a certain level of usage, and you have a reasonable idea of what you will spend over the year. AI is making that much harder. As businesses move beyond individual AI subscriptions and start using AI across more of their operations, costs can vary considerably depending on which models are being used, how often they're being used and what they're being asked to do.

Every era of computing eventually exposes the assumptions that made the last era work. For decades, networking succeeded because it was mostly identity agnostic. Packets moved because they had addresses. Routers and switches forwarded traffic because the network knew where something was going, not necessarily who or what was behind it. That model helped build the internet and modern enterprises. But it is not enough for the era we are entering now.

Ransomware attacks no longer rely on traditional encryption methods. With today's advanced technology, threat actors are developing 'encryption-less extortion', focusing solely on data exfiltration and the threat of leaking or selling stolen sensitive information. Often used as part of double and even triple extortion strategies, ransomware has now evolved into a fragmented, competitive, and increasingly strategic threat landscape that employs divergent attack strategies, laser-focused on high-value targets.

These weren't the test systems Anthropic were looking for.

Network security protects the nervous system of modern businesses. Highly connected and widely distributed throughout the corporate body, it keeps applications regulated, tools performing, and operations stable when functioning well. But like the human nervous system, it is vulnerable to drifting away from an optimal state and becoming weaker, leaving it open to attack. While the human nervous system is affected by stress, lack of sleep, and poor nutrition, digital network security is prone to accidental misconfigurations, ungoverned rule changes, and controls that drift away from their intended state.

OpenAI was running a cyber-capability evaluation against advanced models, including GPT-5.6 Sol and a more capable pre-release model with reduced cyber refusals. The environment was meant to be constrained and the model still found brute forced through it.