Amoreg
Amoreg sells an AI compliance workspace to regulated firms in the Channel Islands, the Isle of Man and the UK. Its buyers are trust and corporate service providers, fund administrators, financial institutions and law firms, along with their MLROs, compliance teams and boards.
The product watches the regulators a firm chooses and checks each new publication against the firm's own policy library. Gaps become findings, each with the rule it breaks and a drafted fix. An assistant called Eve answers questions from the regulator's text and the firm's policies, with citations. Inspector reviews the policy library every week and proposes changes, and Probe rehearses a regulator's visit with examiner style questions.
Agents propose and a named person decides. Nothing Inspector drafts is applied until someone approves it, and every sign off and dismissal is recorded. Client onboarding and screening products are announced as coming soon.
Capability Axes
Capability grades
15 of 15 axes rated · 10 graded A or B
Eve answers compliance questions from the regulator's text and the firm's own policies, and Inspector reads the policy library each week and drafts the fixes. On the monitoring side, an agent reads every watched source, logs each publication and proposes whether it applies to the firm. The findings queue, version history and sign off trail exist to hold what those agents produce. Without them a firm is back to the spreadsheet, shared drive and inbox routine Amoreg sells against.
Agents propose and a named person decides, and the product enforces the split. Inspector drafts fixes but cannot apply them. Only a person can press Apply fix, which saves a new policy version and closes the finding. Eve asks for approval every time it would send an email, decide a finding, edit a policy or change organization settings, and an outgoing email waits until someone approves the exact message.
Sign offs carry a name, dismissals need a written reason, and a finding cannot close without a newer document version or a written note. In monitoring, the view on whether a publication applies is a proposal until a person confirms or overrides it, and the decider is recorded. Scheduled automations, such as a daily sanctions notice, run unattended and report back to the chat.
Every point Eve makes is tied to the passage behind it, either a rule reference or a named policy section, and web sources come with links. Findings name the rule they break, the sections affected and a severity, and anyone can ask Inspector on the finding why it was raised. When a policy changes, open findings are reassessed and fixes the new text already makes are dropped. Amoreg publishes no accuracy or recall figure and shows no confidence signal on answers, and it cautions that its own descriptions are no assurance that every answer is complete or correct.
KPMG, Entrust, Edgewater, Garfield-Bennett / Summit Group, Dominion and Suntera appear as customers. Suntera provides fund, corporate and private wealth administration, and Dominion runs regulated businesses in Jersey, Malta and Switzerland. The one attributed quote, from a compliance officer at Garfield-Bennett / Summit Group, praises meeting minutes drawn from a transcript rather than the monitoring and policy work at the center of the product. No customer puts a figure on time saved or issues caught.
Retrieval is walled by firm. Every database query and vector search runs as the organization asking, a request for another firm's document looks the same as one for a document that does not exist, and retrieval fails closed when scope is unclear. Learned working preferences are scoped just as tightly.
They are short notes on answer format, working steps and standing preferences that belong to one person in one organization, are never used in another organization, and lapse 90 days after they were last confirmed or are deleted when the person leaves. The review that writes them reads what the person typed and short excerpts of the replies they answered, and leaves out documents, search results, tool output and client facts. Each person can view, reword or delete their notes, or switch learning off. Whether customer content ever trains or tunes a model is left open.
Amoreg generally acts as processor for a firm's workspace content, and the firm decides what personal data goes in. It is controller only for its own site, account, billing and support data, and its data protection supervisor is the Office of the Data Protection Authority in Guernsey. Workspace browsers run on Amazon Web Services in London and research browsing on Browserbase in Frankfurt.
Resend sends email and keeps content in the US for 30 days, and Composio, which holds connected app access, may process data in the US. Amoreg gives no retention period for workspace content and does not name its hosting, database or authentication providers.
Amoreg says it has a SOC 2 Type II report and ISO 27001 certification and sends the reports on request by email. There is no trust portal, and the auditor, certification body, audit period, scope and standard edition go unnamed. The engineering controls are concrete.
The application reaches its database over certificate verified TLS and refuses plain connections, every code push runs through secret scanning and push protection, dependency alerts flag security updates, and daily backups are restored in an isolated environment to test them. Encryption at rest and multifactor authentication are not addressed.
Regulators appear in Amoreg as sources to watch. The company does not describe itself as licensed or registered with a financial regulator, and it cites no sandbox or supervised test. It places its agents beside the compliance officer rather than in place of one: monitoring does not replace the officer's judgment, and Probe's rehearsal feedback is informal and developmental, not a regulatory conclusion.
Today the product reads regulations, policies and contracts rather than scoring people, so the bias exposure is limited. Amoreg's controls govern what its agents may do, and it does not describe how the models themselves are evaluated or reviewed. RegBench, its announced benchmark for regulatory reasoning, has no question set, scoring method or results yet. The announced screening product would change that, since matching client names against sanctions and politically exposed person lists produces uneven errors across names and scripts.
Under the public terms, the firm remains responsible for the final compliance decisions its organization makes. The terms exclude indirect and consequential damages, set no cap, carry no indemnity in either direction and make no promise about the accuracy of AI output. A signed agreement or order form overrides them where the two conflict. When an answer is wrong, the trail shows where it came from: every answer cites its source, and every attach, reopen, resolve and close on a finding is logged with who did it.
As of November 2026, Eve's standard chat runs on OpenAI's GPT-6.1 Sol and its Pro option on Anthropic's Claude Opus 5.5. Background work is split by speed, with GPT-6 Astra on the fast worker, GPT-6.1 Sol on the everyday worker and Claude Opus 5.5 on the deep worker, and workspace restrictions can change the route. Probe's voice interviews run on GPT-Live 1, with GPT-5.6 Luna handling lookups and review, and learned preference review uses GPT-5.4 Mini.
Around the models, Pinecone Inference generates embeddings with the llama text embed v2 model, Browserbase runs research browsing, Composio holds connected app access and Amazon Web Services hosts the workspace browsers. Amoreg does not say which platform serves the model calls, where requests run or what each provider keeps, and it notes that the lineup can change.
Documents come in from SharePoint, iManage Cloud and Laserfiche, compliance registers and attestations from RuleGuard, and entity and ownership records from PlainSail, a trust and corporate administration platform. Altruist supplies custodial data as of the prior business day, and Salesforce and HubSpot supply client records. Most connections only read, access is limited to what an admin grants, and iManage follows each person's own permissions. Firms can also plug in a remote MCP server and choose which tools Eve may use. How often each connection syncs is not stated.
Amoreg runs as a cloud service with separate UK and US environments, and Europe and Australia are listed as coming soon. Its production vector store and workspace browsers sit in London. Data may still be processed in Guernsey, the UK, the European Economic Area, the US and elsewhere by Amoreg and its providers, so the UK environment does not by itself keep a firm's data in the UK. There is no on premises or private cloud version.
Pricing comes through a demo rather than a price list, and Amoreg keeps its prices off the site. Most teams start with a pilot on one entity.
The product is packaged separately for trust and corporate service providers, fund administrators, financial institutions and law firms, and for the MLROs and boards inside them. Fund administrators get one workspace per fund and coverage of six fund regimes, from the Jersey certified funds codes and Guernsey's collective investment scheme rules to AIFMD II and SFDR.
For banks, insurers, brokers and fund businesses it works from the FCA's systems and controls rules, operational resilience, DORA, Consumer Duty and the PRA Rulebook. The home market is the Channel Islands, the Isle of Man and the UK, and the regime library reaches into EU, Luxembourg and Cayman rules.
Alternatives to Amoreg
Other products that do the same kind of work as Amoreg, and what each one does.
Sedric turns a regulated firm's policies and the rules it operates under into enforceable system logic, then applies preventive, detective and corrective controls across every customer touchpoint.
Ncontracts sells risk, compliance and vendor management software to banks, credit unions, mortgage companies and fintechs, more than 5,000 of them in the United States.
Sei builds AI agents for customer experience and compliance teams at banks, mortgage lenders and servicers, credit unions and fintechs, handling interactions across voice, chat and email and monitoring them for regulatory issues.
Shiboleth automates consumer lending compliance for banks, fintechs and non-bank lenders, and serves as a monitoring layer for banks overseeing their fintech partnerships.
NuArca Labs sells NuComply, a large language model compliance assistant built for banks and other financial institutions.
Ruleguard is a governance, risk and compliance platform built only for regulated financial services, and it now presents itself as an agentic one.
Pricing
Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.
No pricing data has been verified for this vendor. Pricing information will be published here once confirmed through vendor disclosure or third-party estimation.