Sovereign · India-hosted · Cyber and legal
India needs its own Mythos. We’re building it. Sovereign. For cyber and legal.
Sovereign AI for India. Your data never leaves the country. Built and hosted here, on open-weight models we fine-tune ourselves, for the two jobs that cannot be outsourced abroad: cyber defence and the law.
Access is vetted, NDA-based and for lawful, authorised use only.
Why now
The most capable cyber AI is a foreign, restricted model
In April 2026 Anthropic announced a frontier model, Claude Mythos Preview, and withheld it from public release, sharing it only with selected organisations under Project Glasswing. Indian reporting says the Government has since asked Sarvam AI and BharatGen to build Mythos-like cyber capability on isolated, government-controlled compute. The lesson is plain: critical work needs AI that India controls.
Why sovereignty
Where your data is processed is a security decision
For sensitive work the question is not only how good the model is. It is who else can see what you ask it.
Data residency
The sovereign model runs on our own infrastructure in India. In a sovereign engagement, the models, the documents and the logs stay in India. Nothing is processed offshore.
No foreign-LLM exposure
Most cloud AI sends what you type to large models run abroad. In a sovereign engagement, no prompt, document or output goes to a third-party or foreign language model.
Your jurisdiction, your control
Indian law governs the data, the weights and the logs, under a signed NDA and named people. Nothing is used to train anyone else’s model.
What we are building
A specialist for cyber and legal, not another general chatbot
Open-weight models, hosted here
Built on open-weight models that we host and fine-tune in India, so the weights, the prompts and the outputs never leave our control.
Tuned for cyber and legal
Purpose-built for Indian cyber incidents and Indian law: BNS, BNSS, BSA, the IT Act and the DPDP Act. A specialist, not a general chatbot.
Experts in the loop
Advocates and cyber experts work alongside the models. People decide; the AI assists.
The public record
A credential we can point to
2024 · Public proof
In 2024, our sovereign offensive-security model autonomously solved the full PortSwigger Web Security Academy SQL-injection curriculum and conducted authorized web-application penetration testing.
PortSwigger’s labs are purpose-built training environments, and completing the curriculum is a recognised, verifiable mark of competence, the way a certification is for a person. We state the outcome and nothing more: no payloads and no technique appear on public pages, and every test is run only under written authorisation.
Since 2024 · Classified
Everything since 2024 is classified. It is not disclosed publicly, and is available only to select, vetted key players (defence, government and PSUs, BFSI and select enterprises) under NDA and a scoped, authorized engagement.
Who it is for
Organisations that cannot afford to leak a question
Defence and security establishments
Where the question itself is sensitive and the answer must not leave the country.
Government and PSUs
Departments and public-sector undertakings that need AI within data-localisation expectations.
Banks, financial services and insurance
Regulated institutions handling customer data, fraud and incident work under RBI, SEBI and IRDAI expectations.
Law firms and select enterprises
Chambers and companies that cannot place privileged or commercial material on a foreign cloud model.
How access works
Vetted first, NDA second, scoped third
- 01
Request
Tell us who you are and what you need.
- 02
Verification
We verify the organisation, the requester and the lawful purpose. We may decline without giving reasons.
- 03
NDA
A mutual non-disclosure agreement is signed before any technical detail is shared.
- 04
Scoped engagement
Access, where granted, is limited to a written scope, named users and an agreed period.
Questions
Straight answers
Is this Anthropic’s Mythos?
No. Mythos is a product of Anthropic and we are not affiliated with it. “India needs its own Mythos” is our mission, not a product name and not a claim of equal capability.
Does my data leave India?
Not in a sovereign engagement. The models, documents and logs stay on infrastructure we operate in India, and nothing is sent to a third-party or foreign language model. This describes sovereign engagements only; our free consumer WhatsApp service is a separate product.
Which models do you use?
Open-weight models that we host and fine-tune in India. We do not resell a foreign model.
Can it run inside our own environment?
Yes. On-premises and air-gapped deployment at your site are offered deployment options. The sovereign model itself runs on our own infrastructure in India.
Are you certified or empanelled?
No certification or empanelment is claimed on this page. If one applies to your procurement, ask us and we will answer in writing.
Please note. Access is vetted, NDA-based and for lawful, authorised use only. We do not accept requests for unauthorised testing, for surveillance of individuals, or for any use that would contravene Indian law. Submitting the form does not give access to anything.
Mythos is a product of Anthropic. Codesnag and VirtualVakil are not affiliated with, endorsed by or powered by Anthropic or Mythos, and we make no claim of benchmark parity with it. The name is used here only to describe the national conversation.