Sepanta launches AI governance system for regulated firms

SEPANTA has launched GATE, an AI governance operating system for regulated businesses, aimed at a widening oversight gap as more firms use AI in decision-making.

The launch comes as Bank of England research points to uneven understanding of AI in financial services, despite widespread efforts to assign responsibility for its use. The central bank found that 84% of UK financial firms have named someone accountable for AI, while 46% only partially understand the AI they use.

The research also identified data-related issues among the main perceived AI risks, including privacy and protection, quality, security, and bias and representativeness. With 55% of AI use cases in financial services involving some degree of automated decision-making, questions of control and auditability have risen up the agenda for regulated companies.

GATE is intended to create a record of AI-assisted decisions that cannot be altered after the event. The system records who approved a decision, which policy was applied, and what evidence informed the outcome, while continuously checking controls and flagging departures from approved policy.

The focus reflects a broader debate in regulated industries over how organisations document AI's role in decisions that affect customers, compliance, and operational risk. In sectors such as financial services, firms face pressure to show not only that a person is accountable for AI use, but also that internal policies and regulatory requirements are reflected in day-to-day processes.

Governance gap

UK regulators and central banks have paid growing attention to AI governance as firms move beyond trials into operational use. The Bank of England's findings suggest formal accountability structures are becoming more common, but understanding of the underlying systems remains uneven.

For businesses, that creates a practical problem. If decision-making is assisted by AI systems, companies may need to show how a decision was reached, whether the system was used in line with internal rules, and whether controls remained effective over time.

SEPANTA said its system is designed to address this by embedding governance into the workflow rather than treating oversight as a separate reporting task. It described GATE as a mechanism that sits around AI-assisted decisions, rather than simply producing reports after decisions have been made.

Omid Pakseresht, Chief Executive Officer of SEPANTA, said the issue should be addressed before AI tools are widely deployed inside businesses.

"AI adoption is growing rapidly in enterprises but CEOs need to ask who or what is governing the AI? AI governance cannot be something organisations bolt on after deploying a model. When AI is already making or influencing decisions across businesses, particularly regulated industries such as financial services, governance needs to be built into the technology from the outset with clear accountability, transparency and human oversight," said Omid Pakseresht, Chief Executive Officer, SEPANTA.

Audit trail

One of GATE's core features is the creation of an auditable record for each AI-assisted decision. This includes approval details, policy references, and supporting evidence, alongside ongoing checks intended to identify drift from approved rules as it happens.

Such features speak to a recurring concern in regulated sectors: organisations often have governance frameworks on paper, but can struggle to connect them to live systems and business decisions. The challenge grows where AI outputs influence actions across multiple teams or functions.

SEPANTA, which is based in the UK, said it also uses the product internally. The approach is intended to ensure decision records and control checks are available before clients or regulators ask for them.

Pakseresht said firms can no longer afford to treat governance as a secondary issue.

"Effective governance is key to successful AI adoption and organisations can no longer treat governance as an afterthought. We practice what we preach: everything we build runs under GATE's oversight, including checks and a full decision record, all living under our own delivery. So the evidence is ready before anyone asks for it. Simply put, we govern AI before it acts, not after the fact," said Pakseresht.