When AI Agents Act on Your Behalf, Who’s Actually in Control? AI systems don’t just answer questions anymore, they act. They check your calendar against availability and make the booking. They read your inbox and reply on your behalf. Each time, that agent is operating under its own digital identity, with its own credentials and permissions, separate from you as a person. Security researchers call this a “non-human identity”, and agentic AI doesn’t just multiply how many exist, it changes what they are: unlike a static API key or service account, an agentic identity can reason about what access it needs, request new permissions, and act across systems on its own. This creates a real risk with a real name: the “confused deputy” problem: a privileged system tricked into misusing its own authority. An agent can be manipulated by hidden instructions buried in an email, document, or webpage it reads as part of a routine task, and because it’s using access it legitimately holds, traditional security tools often don’t flag it. In 2026 alone, this played out for real: a Microsoft Copilot Studio vulnerability (CVE-2026-21520) enabled data exfiltration through indirect prompt injection, and a supply-chain compromise of LiteLLM exposed credentials tied to an estimated 500,000 corporate identities. There’s a related but distinct risk too: recently, X and OpenAI separately disclosed a China-linked bot farm using AI to generate content aimed at turning American opinion against data centre construction; humans set the strategy, AI generated the personas and content at scale. Different threat model, same lesson: as agentic AI multiplies, identity becomes the front line. Morgan Stanley’s research team, on its Thoughts on the Market podcast, estimates agentic identity alone could become a roughly $33 billion global opportunity, pushing the broader identity security market past $60 billion in the coming years. Their reasoning: organisations are adding hundreds of machine and AI identities for every human employee, and today only around 39% of privileged access runs on “zero standing privilege” — access granted just-in-time and revoked the moment a task ends. Closing that gap at machine scale is the opportunity. Here’s what doesn’t get discussed enough: securing agentic identity isn’t just a software problem. It needs real infrastructure — credential brokers, hardware security modules, continuous real-time monitoring, immutable audit trails — all running with the resilience and uptime that mission-critical infrastructure demands. As AI agents move from pilots into production across banking, healthcare and everyday services, the data centres hosting them will need to host this identity and security layer alongside the AI workloads themselves, not bolt it on later. Agentic AI is going to keep making decisions on our behalf. The question is whether those decisions stay ours.

Turkey Is Betting Big on Becoming a Data Centre Hub Data centres aren’t just commercial real estate — they’re mission-critical infrastructure. In plain terms: every bank transfer, hospital record, streaming service and AI query you rely on is running on a server sitting in a data centre somewhere. If that facility goes down, the services built on top of it go down too — which is exactly why not just the industry, but governments and official bodies increasingly treat data centres the same way they treat power plants or water systems: essential, not optional. Turkey is treating them that way. The new programme commits up to $30 billion in high-tech investment support by 2030, targeting 1GW of data centre capacity. This year alone, the government has allocated $3 billion to data centre and AI investment, while working to raise a further $10 billion in private capital. As @IDCAorg International Data Center Authority (IDCA)’s exclusive regional representative for Turkey, @IsbasaranGroup Isbasaran Group sits at the centre of this build-out — and we’re involved on more than one front. We invest in data centres directly, we develop and build them, and, as IDCA’s regional partner, we also deliver IDCA’s training and certification programmes and auditing along with design and engineering — helping build the qualified workforce the industry needs, not just the buildings. Having also explored data centre opportunities in the UK, including submitting for a UK tender, the contrast is striking. A recent Computer Weekly piece on UK datacentre planning (Patel & Pauncefort-Duncombe, BTG Eddisons, 24 Aug 2026) lays out a market constrained by planning delays and grid bottlenecks — while Turkey moves with real capital, real incentives and real urgency. The AI infrastructure race will be won by whoever builds fastest and most credibly. Turkey looks ready to compete.

RT @UKaanIsbasaran: Londra’daki Türkiye Büyükelçiliği Rezidansı’nda, T.C. Cumhurbaşkanlığı Yatırım Ofisi ev sahipliğinde düzenlenen kapalı…

It was a pleasure to attend the Republic Day celebration hosted by His Excellency Ambassador Koray Ertaş and Mrs Sevcan Ertaş. Inspiring to see leaders and professionals from Türkiye and the UK come together to celebrate our shared pride and deepen our ties. 🇹🇷🇬🇧

At @IsbasaranGroup , we’re proud to build thoughtfully designed, environmentally responsible homes that meet the needs of real communities. Grateful to see this important conversation at @UKParliament and proud of our commitment to create meaningful impact.