If you’ve been following the ongoing debate around AI safety, you’ve likely seen the leaders of the major AI labs, Anthropic, OpenAI, xAI, talking about the need to “pace the frontier” and build better guardrails around autonomous, self-improving models.

When the people building the most powerful tech in the world call for a breather, it naturally raises a few questions. And if you work in local government, healthcare, or housing, where trust, accountability, and data protection are non-negotiable, those questions carry even more weight.

As CTO, people often ask me how these industry-wide debates affect what we’re building at Orlo.

My take? It simply reinforces that we chose the right path from the start.

While frontier labs grapple with how to safely manage unconstrained, fully autonomous models, our focus at Orlo is completely different. We aren’t trying to build autonomous super-intelligences. We build practical, targeted AI designed to solve real operational problems for public sector teams.

Here is how we look at it:

Much of the wider concern in tech right now centres around AI operating independently without oversight. In our world, AI is never in the driver’s seat by itself.

We design tools that handle heavy administrative lifting, summarising citizen conversations, routing incoming messages, triaging urgent issues, or suggesting draft responses. The goal is to give human staff better context to make decisions faster, keeping a human in the loop every time.

A few months ago, we published our stance on the Responsible Use of Artificial Intelligence. We didn’t do it to react to a sudden news cycle; we published it because public sector organisations carry a massive duty of care. Governance, transparency, and safety can’t be an afterthought you patch in later, they have to be baked into the system architecture from day one.

Take Shield and Assist, built directly with a UK police force: the grading structures, escalation triggers and evidential trail were designed into the system from day one, not layered on after the fact. That’s the same principle the UK Government’s own AI Playbook asks public sector teams to hold their suppliers to: meaningful human control at every stage, not just when something goes wrong.

Data privacy remains one of the largest friction points in the AI ecosystem. At Orlo, our position is straightforward and non-negotiable: client data is protected, isolated, and never used to train public or frontier AI models. You maintain total control over your compliance and security posture.

It’s easy to get swept up in the big debates about where frontier AI is heading over the next decade. But while the underlying model providers work through safety standards and governance, our promise to you stays the same:

We’ll keep delivering safe, transparent, and purposeful tools that make public sector work easier, keep data secure, and keep public servants squarely at the heart of public service.