ComplyDNA is a compliance LLM trained on Turkish regulation. It knows the MASAK communiqués, the AML/CFT laws and KVKK with their chain of sources, and answers every sentence with an inline [COMMUNIQUÉ / Article] citation. Auditable answers, not guesses.
Every claim arrives with a [COMMUNIQUÉ / Article] tag. Which
piece of regulation an answer rests on is traceable at sentence level — the
compliance team verifies the source in one click.
MASAK communiqués and circulars, laws 5549 and 6415, KVKK, the ROM Regulation and crypto-asset rules are in the model's training set — no translation layer, regulatory language handled directly.
Answer + source list + version info are produced together. Thanks to effective-date metadata, "which provision applied on which date" can be answered too.
ComplyDNA is not a RAG wrapper you upload PDFs into and chat with. Every stage is designed around a compliance team's audit requirements: article-level metadata, a retrieval-confidence threshold, mandatory citations and version traceability are part of the pipeline — not a check bolted on afterwards.
Raw communiqué, law and regulation texts are converted into structured article/paragraph records.
source_code, article_no, paragraphvalidity_status, effective_date, versionArticle chunks are written into a semantic index; queries can be in Turkish or English.
The model learns regulatory language and the citation format; uncited output cannot pass the pipeline.
[SOURCE / Article X] reference format and regulatory TurkishRetrieved article texts + the fine-tuned model = an auditable output package.
[COMMUNIQUÉ / Article]; enforced by post-check> query: Within what period must a suspicious transaction be reported?
Suspicious transactions are reported to the MASAK Presidency within ten business days at the latest from the date the suspicion arises — and immediately where any delay would be risky [MEASURES-REG. / Article 28]. The reporting obligation arises from the institution's obliged-party status under Law No. 5549 [LAW-5549 / Article 4].
--- SOURCES ---
[1] MEASURES-REGULATION / Article 28 · validity: in force
[2] LAW-5549 / Article 4 · validity: in force
ComplyDNA’s primary product is cite-first Ask. The modules below are design-partner scope — not generally available; methodology details are not published publicly.
Lower / Medium / Higher bands for onboarding and corridors — multi-source AML geographic risk alongside sanctions and FATF lists. Does not replace customer due diligence.
Coming soon · design partnerA VASP / fintech compliance frame that ties Board-set appetite limits to EWRA and KRI monitoring — accept, restrict, and prohibit rows.
Coming soon · design partnerMASAK.ONLINE-ready ŞİB XML draft and validation support. The compliance officer reviews and uploads — no auto-submit.
Coming soon · design partnerDesign-partner pilots discuss suite scope alongside Ask. Module methodology is shared only under NDA / pilot terms.
Prices in USD. Team is billed annually. Enterprise starts at $1,990/mo and scales with regulatory corpus scope, seats, and deployment (cloud or on-prem). Questions: hello@ozdna.com.
ComplyDNA is a compliance LLM trained on Turkish regulation (MASAK communiqués, AML/CFT laws, KVKK). It answers compliance questions with an inline [COMMUNIQUÉ / Article] citation on every sentence — the piece of regulation behind each answer is always verifiable.
No. Ask is the primary product. Those three modules are on the roadmap and discussed only in design-partner pilots; they are not generally available.
Law No. 5549 on the Prevention of Laundering of Crime Proceeds, Law No. 6415 on the Prevention of the Financing of Terrorism, the Turkish data-protection law KVKK (6698), MASAK communiqués and circulars, the Measures Regulation, and crypto-asset service provider (VASP/CASP) rules. Coverage is versioned and updated as regulation changes.
General models can't cite sources on Turkish regulation questions and don't ground answers in current communiqué text. ComplyDNA builds each answer by first retrieving the relevant article text, ties every claim to its source, and writes in regulatory language. Quality is measured on every release against a golden set written by compliance experts.
No. Customer queries and documents never enter model training. On the Enterprise plan the model and index run entirely inside your own infrastructure; data never leaves your perimeter.
Both: a fast start via cloud API / web interface, or a fully on-premises installation for banks and financial institutions. Pilot programs usually start in the cloud.
No. ComplyDNA outputs are informational and do not constitute legal advice; final decisions require verification by a compliance professional. The inline citations exist precisely to speed up that verification.
The early access program is open. We prepare a demo tailored to your institution's regulatory scope — one email is enough.