01
Understand AI
Study how AI learns, reasons and acts.
LiveFreedom v1.1 — India’s largest cybersecurity LLM
Securing the age of intelligence.
We research how AI works, secure how it acts, and build systems that enterprises can rely on.
Program
- NVIDIA Inception
Engagements
- IndiaMART
- Angel One
Research
- IIT Madras
- The University of Texas at Austin
- Walsh College
- IIM Ahmedabad
- IIM Raipur
- IIT Kanpur
- Freedom v1.1, total logical parameters
- 602B
- Benchmarks where Freedom v1.1 places first
- 8 of 12
- Scenarios across 51 cyber domains
- 431
Why Cosmipher
Understand AI. Secure it. Build with it.
The same understanding that helps us find weaknesses also helps us build better AI systems.
02
Secure it
Protect every path an agent can take.
03
Build with it
Put the same foundations to work.
Engagements and programs
We work with enterprises to secure their systems
Programs and engagements show how we apply AI security and governance to real workflows.
- EngagementIndiaMART
- EngagementAngel One
- ProgramNVIDIA Inception
- ProgramOpenAI Daybreak Blue
- ProgramGoogle
- ProductAdReview
Enterprise engagement
IndiaMART
We work with IndiaMART to secure its agentic systems, assess exposure and design controls for agent actions.
Enterprise engagement
Angel One
We built AdReview to screen advertisements against compliance requirements and route evidence to reviewers.
pilotThe pilot focuses on the review workflow.Program
Backed by NVIDIA Inception
Eligible startups can access ecosystem offers worth nearly USD 300,000 in credits and discounts.
Program
OpenAI Daybreak Blue
A member of our team has Daybreak Blue access for responsible cybersecurity research.
Program
Google
Our agents run on Google infrastructure. Our researchers responsibly disclose AI vulnerabilities to Google.
Freedom v1.1
India’s largest cybersecurity LLM
Freedom v1.1 is built in India for cybersecurity research, authorized testing and enterprise defense and offense.
602B
Total logical parameters
5.73×
Parameter count relative to Sarvam 105B. Nearly 6× as many parameters.
Mixture of experts
Nominal active parameters: 8B during prefill, 16B during decoding
Multimodal image and text input
Input
Benchmark highlights
Freedom v1.1 beside its strongest peer on four cybersecurity benchmarks.
CyberGym
CyberGym leaderboard / one-trial model track
CTIBench RCM
QCRI audit / 2026
CyberMetric 10K
QCRI audit / 2026
SecEval
QCRI audit / 2026
Security Capability Atlas
97.00%
20-case Security Showcase
92.34%
431-scenario full catalog. 431 demanding scenarios across 51 cyber domains.
Gold-standard data
~400 TB
Compressed cybersecurity knowledge. All five Vs of big data.
Where it goes to work
- Vulnerability research and secure-code review
- Threat intelligence and SOC investigations
- Cloud and identity security
- Malware and binary analysis
- Security testing for agents, MCP and retrieval systems
Mixture of experts8B during prefill, 16B during decoding
Research
Continual Systems Intelligence
A shared coordinator routes work to the agent harness best equipped for the task.
Govern
Intent, risk, permissions and promotion rights are fixed before any work begins.
Execute
Agents work under that contract, and every trajectory is recorded with its evidence.
Assure
Independent assurance judges each outcome. Model confidence never stands in for evidence.
Compile
The Experience Compiler promotes repeated verified reasoning into the lightest reusable form: an episode, a procedure, a skill or a deterministic tool.
Experience accumulation
Useful successes and failures keep their provenance, so the system never starts from zero.
Experience compilation
Repeated verified reasoning migrates to cheaper reusable capability once held-out evidence supports it.
Evidence governance
Every active artifact stays scoped, provenance-linked and revocable. Contradicted evidence re-evaluates everything built on it.
Research collaborations
- IIT Madras
- The University of Texas at Austin
- Walsh College, Michigan
- IIM Ahmedabad
- IIM Raipur
- IIT Kanpur
Targets
What we are building toward next.
80%
Token-cost reduction target
40×
Speed-up target
1,000s
Parallel-agent capacity target
Secure it
Protect every path an agent can take.
A model can retrieve data, call tools, delegate tasks and change business records. Each connection adds a security boundary.
The AI security platform
The wider catalog expands these roles into 32 tools and modules across the two groups below.
01
AgentSPM
Discover models, agents, identities, tools and data connections.
Explore on the AI Security site02
AI Firewall
Evaluate interactions and tool actions at supported boundaries, then apply the policy and retain a reviewable decision.
Explore on the AI Security site03
AI Security Testing
Exercise authorized failure paths, preserve reproducible findings, guide remediation and replay the original attack after a change.
Explore on the AI Security site04
AI Supply Chain
Connect exact artifact identity, composition, lineage, inspection, ownership and release approval.
Explore on the AI Security site
AI security and cyber defense tools
D01Protect AI
- MCP Guard
- Agent IAM
- AI-BOM Registry
- AgentForensics
D02Use AI for defense
- Sentinel
- Hunter
- Vulcan
- Forensicator
Market
The market for securing AI
Our focus is the agent lifecycle.
IDC
$409B
Global AI IT spending projected for 2026 (IDC)
Spending to secure AI
Gartner forecast / USD billions
- $2.835B2026
- $4.783B2027
- ~$7.7B2028
The estimates cover different categories.
Direction
Mission, vision and direction
Mission
Secure the age of intelligence by helping organizations deploy and operate AI they can rely on.
Vision
Build AI in India that enterprises worldwide can trust with their most demanding work.
Where we go next
Deepen the research
Use the model, the cybersecurity corpus and the capability atlas to test security capability with stronger evidence.
Extend agent protection
Connect discovery, testing, runtime policy and release evidence across more enterprise systems.
Reuse what works
Carry shared context, permissions and verified experience into the applications we build. Use what we learn there to improve the underlying systems.
We measure progress by what teams can use securely.
Securing the age of intelligence.
Ciphering the cosmos means understanding how models, agents and systems work so we can secure them and build AI people can rely on.