AI Is Scaling Faster Than Our Data Security Assumptions

February 23, 2026

As AI Scales, can Data Security Keep Pace?

A recent AI macro deck from Andreessen & Horowitz has been circulating widely among security and technology leaders. It is a strong piece of work, particularly in how it frames the economics of AI, the scale of hyperscaler investment, and where long-term value is likely to accrue.

What stood out most was not just what the deck stated directly, but what it quietly assumed.

As AI adoption accelerates, there is an implicit belief that the underlying infrastructure will scale alongside it. Especially when it comes to data movement and encryption, many assume this layer will simply keep up. In regulated and high assurance environments, the assumptions underpinning AI and data security at scale are already being tested

AI Changes How Data Moves, Not Just How It Is Processed

One of the clearest signals in the deck is the sheer scale of what is coming. Trillions in AI driven revenue, unprecedented infrastructure spend, and AI embedded across everyday workflows rather than isolated use cases.

What follows from this is often under discussed. AI dramatically increases data in motion.

Not just traffic between users and systems, but continuous machine to machine flows, an increasingly AI driven data in motion pattern. Data moving between data centres, across sovereign and regulatory boundaries, from edge to core to cloud, and between autonomous services acting on each other’s behalf. In many environments, these flows are persistent, high-volume, and increasingly latency sensitive.

This shift matters because most security models were designed for a very different world, one where data movement was more bounded, more predictable, and easier to contain

Where the Friction Starts to Show

In highly regulated sectors such as banking, financial services, government, and critical infrastructure, teams are already feeling the pressure created by AI driven data in motion. These are environments where performance, availability, and security are tightly coupled, and where failures tend to be systemic rather than isolated.

Latency becomes a hard constraint. Encryption overhead is quietly traded off for performance often in ways that are difficult to see from a policy or audit perspective. Compensating controls are layered on top of fragile assumptions. Gaps open up between compliance intent and operational reality.

This does not mean teams are failing at security. It means the system boundaries have moved, and security controls are now operating in places and at scales they were never originally designed for.

As AI scales, the transport layer becomes a critical part of the risk surface. It is also a layer where existing security stacks are being stretched beyond what they were originally designed to handle. For many organisations, this is unfamiliar territory, a layer that was previously assumed to be stable, invisible, or someone else’s problem.

Infrastructure Is Where Durability Lives

One of the strongest themes in the deck is that AI winners will not all be visible applications. A disproportionate amount of long-term value will sit in infrastructure. The layers that are hard to replace, deeply embedded, and quietly assumed.

This is especially true in environments where performance is non-negotiable, compliance is continuous, and failures are systemic rather than local.

In those contexts, encryption cannot be a performance bottleneck, an operational burden, or a future cryptographic risk. It must operate at line rate, behave deterministically, and remain manageable without becoming another system teams need to fight.

Why This Matters Now for Regulated Industries

For sectors like banking, financial services, and insurance, this is not a future concern. It is already present in payments and clearing infrastructure, trading and market data environments, interconnects between regulated entities, and early sovereign and private AI deployments. These environments already operate at scale, under tight latency and availability constraints, and with little tolerance for unpredictable behaviour.

As AI increases both the volume and value of data in motion, the cost of getting this layer wrong compounds quickly, across performance, resilience, and compliance.

If encryption becomes the bottleneck, AI return on investment collapses.

A Closing Thought

The deck making the rounds does an excellent job of explaining why AI investment is accelerating and where economic value is likely to concentrate.

The next question for security and technology leaders is simpler, and harder.

Are the assumptions we are making about data in motion still valid at AI scale?

It is not a question with a single answer. But it is one worth asking early, before performance, security, and compliance start pulling in different directions.

Take a demo of Sitehop SAFE Series today!

To find out more, email info@sitehop.com

Or call us: +44 (0)114 478 2366

Sitehop.

Engineered for speed. Built for the future.

About Melissa Chambers – CEO & Co-Founder, Sitehop

Melissa is the CEO & Co-Founder of Sitehop, and leads the development and scaling of high-speed, post-quantum encryption hardware and securing data-in-motion without compromising performance. With a background in hardware engineering, Melissa specialises in designing, manufacturing, and scaling deep tech products from concept to market.
Passionate about solving complex problems through engineering and advocating for women in STEM, Melissa has seen firsthand the impact of diverse technological perspectives. A proud member of Cyber Runway Ignite and the Women’s Engineering Society, Melissa was honored to be recognised as Startup Magazine’s 2023 Inspirational Woman in Industry and 2024’s Most Inspiring Woman in Cyber, as well as leading Sitehop’s win in the recent ClimbUK Awards with Cyber Security Innovation of the Year 2025.