Cabinet presents AI vision and explores a Dutch AI safety institute
On 6 November, the Council of Ministers approved the cabinet's AI vision put forward by State Secretary Willemijn Aerdts. The cabinet wants to remove barriers for SMEs, jointly procure AI applications for government, and create room for supervised experimentation. Healthcare, energy, food production, security and high tech are the priority areas, and from 2027 onwards a plan for the necessary computing power will be developed.
What does this mean? If you work in government or healthcare, expect joint procurement and testing grounds for AI in the coming years rather than individual choices made by each organisation.
NCTV: attackers use AI at every stage of a cyberattack
According to the Dutch Cyber Security Picture 2026 from the NCTV, attackers are using AI to prepare attacks and carry them out on a larger scale, including for malware and more convincing phishing. The digital resilience of Dutch organisations is lagging, especially among SMEs. The NCTV advises good patch management, logging and monitoring, and including AI risks in security policy.
What does this mean? Include the safe use of AI tools in your security policy and expect phishing emails to become ever harder to tell apart from genuine ones.
Microsoft limits the free Copilot Chat for work accounts from 9 November
Microsoft is introducing usage limits for Copilot Chat for users without a Microsoft 365 Copilot licence in business environments, on web, desktop and mobile. The rollout starts on 9 November. Microsoft does not disclose figures: the limits may vary by time of day, and administrators cannot view or adjust them.
What does this mean? If your team relies on the free Copilot Chat, bear in mind that it may refuse requests at busy times, and check whether a licence or another model is needed.
OpenAI releases GPT-6 Cyber to a limited group of security teams
On 5 November, OpenAI released GPT-6 Cyber, a model for authorised vulnerability research. According to OpenAI, it found 81 new vulnerabilities in Firefox and developed exploits for 948 vulnerabilities in the Linux kernel. The model is initially available only to a limited number of customers in the Daybreak Red programme, with additional identity checks and monitoring.
What does this mean? For most organisations the model is not directly usable, but it shows how quickly AI can find vulnerabilities: fast patching becomes even more important.
EU AI Office selects Capgemini-led consortium to test cyber risks
The AI Office of the European Commission has selected a consortium of Capgemini, Wavestone and the French testing laboratory LNE. The consortium will develop threat models, evaluation tools and methods to verify compliance, and will carry out technical assessments of advanced general-purpose AI models under the EU AI Act.
What does this mean? Oversight of the large models you use every day is getting technical teeth, which may lead to greater openness about their risks over time.
