SAIE for banks and insurers is a layer for controlling and working with AI in which every AI-supported decision has an author, data and the audit trail required by supervisors. AI under KNF supervision: every decision has an author, data and a trace
Why now
Banks and insurers must show who authorised an AI-supported decision and on what data. DORA, the AI Act and ISO/IEC 42001 overlap, client data must not leave the organisation, and clients test the limits of models in service channels. A pilot that cannot be defended before the supervisor will not reach production. On 22 July 2026, UKNF, the office of KNF (Polish Financial Supervision Authority), issued recommendations on Frontier AI models, and in its priorities for 2026 it stated that it will itself analyse institutions’ reports with the help of AI. From 28 October 2026, KNF is part of KRiBSI (Commission for the Development and Security of Artificial Intelligence), which oversees the application of the AI Act. Since 17 January 2025, DORA has required AI providers to be included in the register of ICT contractual arrangements. Artificial intelligence in banking and insurance is moving beyond pilots: according to an EY survey (April 2026), 52% of financial sector companies have completed their first AI deployments, and regulatory constraints are their biggest barrier (43%). Banks are also covered by the amended KSC Act and apply for entry in the register of essential and important entities, but DORA takes precedence in ICT risk management and incident reporting
Business scenarios
Budgets and AI cost control
Cost and efficiency
PROXY:AINEW · SAIE PANTERAProblem: Spending on language models is unpredictable and not assigned to an owner
How it works: Hierarchical budgets per team and project, with a policy for when the limit is exceeded: block, downgrade the model or raise an alert. The Executive AI Dashboard shows the board costs, subscriptions and the forecast on one screen
What you gain: Every zloty spent on AI has an owner and a limit, and exceeding the budget triggers a predefined response
Document analysis with source verification
Knowledge and documents
TWIN:DESKNEW · SAIE PANTERAProblem: Analysing contracts, applications and claims files requires certainty about the source of every piece of information
How it works: A central RAG knowledge base with source preview and a Knowledge Gap Register; OCR of scanned documents. Agentic RAG asks again in another round when something is missing, and TWIN:INDEX searches long documents by chapter
What you gain: The analyst sees which document each piece of information comes from, and the system says plainly when the file lacks a basis
Safe AI work with client data
Security and sovereignty
PROXY:AIProblem: Client data must not reach a model outside the organisation
How it works: Model routing based on data sensitivity, detection and masking of personal data, and a predefined Finance sector profile
What you gain: Staff use the best models for general work, and client data goes only to the local environment - automatically
Protecting models against manipulation and abuse
Security and sovereignty
PROXY:AIProblem: Clients try to manipulate the model in service channels
How it works: Real-time abuse detection: injection, jailbreak, prohibited content and token-usage anomalies
What you gain: An AI service channel that resists attempts to extract data or bypass the rules - and every attempt leaves a trace
A full audit trail of AI-supported decisions
Compliance and oversight
PROXY:AI + TWIN:DESKNEW · SAIE PANTERAProblem: The supervisor requires evidence of who authorised an AI-supported decision and on what data
How it works: A cryptographically signed, immutable log of every call (Article 12 of the AI Act) with retention, and a Human Oversight mechanism (Article 14) with approval gates. Amounts come from a decimal calculator with a calculation card, and a report with a mismatched figure is not published
What you gain: The answer for the supervisor is in the system: who, when, on what data and with what result - without reconstruction
Informed approval of AI-supported decisions
Compliance and oversight
CDFProblem: An employee approves an AI recommendation without really checking it
How it works: Human Competence Gate (CDF-F2) - before a critical decision is approved, the system asks 5-15 check questions; too few correct answers block approval
What you gain: Human oversight stops being a “click approve” fiction - and you can prove it to the auditor
Inventory and risk classification of AI systems
Compliance and oversight
PROXY:AI + CDFProblem: The regulator asks for an inventory of AI systems and their risk classification
How it works: CDF-F1.5: AI Act readiness assessment, statement of applicability and impact assessment report; PROXY:AI: a register of models and policies and the decision history as material for ISO/IEC 42001
What you gain: An inventory of AI systems with risk classification, ready for the supervisor’s request - kept up to date, not built for an inspection
Questions decision-makers ask
How can you limit hallucinations in banking applications?
By answering only from approved documents, with a source citation, and through human oversight of the decision. TWIN:DESK answers based on the organization’s documents with a citation, and PROXY:AI lets you route tasks to models proven in a given use case
How do you assess an AI provider and avoid concentration?
By being able to swap the model without rebuilding the process. PROXY:AI provides a single gateway to more than 100 models, both cloud and local. Changing a model is a configuration change, not a new deployment
Bielik and PLLuMHow do you include an AI provider in the register of ICT contractual arrangements (DORA)?
Like any ICT service provider supporting important functions: with a risk assessment, an exit plan and information on data location. The WORM log in PROXY:AI documents which models processed data and when
What KNF guidelines apply to artificial intelligence?
The most recent are the recommendations of UKNF (Office of the Polish Financial Supervision Authority) of 22 July 2026 on the impact of Frontier AI models on cybersecurity. They build on DORA and recommend, among other things, taking AI-enabled attacks into account in ICT risk analysis, assessing whether ICT providers manage AI risk, and reducing the risk of over-dependence on a single ICT or AI provider. UKNF has announced that it will verify their implementation as part of supervision. PROXY:AI helps meet them: the WORM register records every model call as evidence for the supervisor, and a single gateway to more than 100 models reduces dependence on one provider. Since 17 January 2025, DORA has replaced Recommendation D and the earlier KNF IT guidelines
All questionsOther paths: Finance and professional services
AI for lawyers with no invented case references and no case files in someone else’s chatbot
AI that reads invoices, tracks deadlines and keeps client data separate
Faster deliverables, expert knowledge retained in the firm, client data kept separate