Moyo logo
Buy a report

Methodology

How it works

Moyo does not break into systems. It asks the same kinds of questions an adversary would ask of public large language models (LLMs) like ChatGPT and Deepseek, then turns the answers into scored findings with real-world citations. Use it on your own footprint, or to learn what is already reachable about a competitor, adversary, or opponent.

Moyo only probes what models can already infer about your topic from public information. Do not paste secrets, credentials, or private documents. Need the scan to run under your control? Ask about licensing.

See pricing & buy a report Exposure Data, Snapshot & Basis Report →

Who it’s for

Anyone who cares about sensitive information

If you need to see what models can assemble about you, a competitor, an adversary, or an opponent from public sources, Moyo is built for you, especially:

  • · CISOs and security leadership who need a board-ready read on outside-in exposure
  • · Corporate counsel, privacy, and compliance teams tracking what models can infer from public records
  • · Defense, intelligence, and public-sector OPSEC programs
  • · Political campaigns, parties, and public figures managing reputational and OPSEC exposure
  • · Opposition research and political intelligence teams mapping what models can assemble from open sources
  • · Product, GTM, and communications teams before launches or partnerships
  • · M&A and diligence advisors mapping inferable strategy without claiming private access
  • · AI governance owners of RAG corpora, assistants, and customer-facing models
  • · Companies with competitive, regulated, or classified-adjacent intellectual property
  • · Competitive intelligence and strategy teams mapping a rival from public sources
  • · Founders and operators who want to know what the internet already reveals about them or about an opponent

Process

Four steps

  1. 01

    Start, scope & explore

    Investigate a topic first with a free Exposure Data scan (one topic, about 15 minutes). Additional Exposure Data scans are paid. Or buy an Exposure Snapshot or a Basis Report. Snapshot and Basis: after payment you name the organization and may add points of exploration (up to 3 for Snapshot, up to 10 for Basis), or choose Let us guide the investigation. No credentials. No private uploads. Outside-in only.

  2. 02

    Probe models

    Our analysis worker queries multiple frontier models with varied strategies against lawful public sources (filings, news, disclosures, jobs, posts, and other open material), using your points of exploration as the investigation brief.

  3. 03

    Extract, score & QC

    Human QC gates packaged Snapshot and Basis delivery. Exposure Data skips that queue: extracted claims and structured findings go out when the scan finishes, typically about 15 minutes. The first scan is free.

  4. 04

    Deliver & act

    Exposure Data: first scan free, then extracted claims and structured findings in about 15 minutes — use AI as you see fit to crunch and present. Snapshot or Basis Report: packaged for presentation, delivered by email or secure link, typically within 48–72 hours after scope is confirmed, with a clear path to remediation, re-test, or authorized white-box work.

Output

Models corroborate claims. Claims support conclusions.

Independent models each extract candidate facts. When several models agree, that fact becomes a claim. Corroborated claims then support chains: the conclusions an outsider can already reach. The graph below is from a real Moyo run.

Evidence graph · Models → claims → chains
Claude DeepSeek Gemini Llama-3 Mistral Perplexity GPT-4 Qwen Phi-3 C0040 C0038 C0047 C0056 C0048 C0046 C0052 C0045 C0050 a new rocket core C0058 a new 30kN engine C0054 a 200kg payload C0062 pilot hiring requirements MODELS CLAIMS CHAINS
  1. 01

    Models

    Multiple systems independently read the same public footprint. Real-world citations stay attached.

  2. 02

    Claims

    Agreement across models turns a candidate fact into corroborated evidence.

  3. 03

    Chains

    Those claims support conclusions: a capability, a timeline, a hire plan.

Topic scan

What do AIs know?

Run a free Exposure Data scan for one topic in about 15 minutes, or buy a Snapshot or Basis Report packaged for presentation.

Investigate a topic

Products

Three outside-in products, one analytic pipeline

All three are public-source only. Exposure Data is one topic, fast — use AI as you see fit to crunch and present. The Snapshot answers what stands out and is packaged for presentation. The Basis Report answers what exactly was found, how models reached it, which sources they cited, and what to do next — packaged for presentation.

  1. 01

    Exposure Data

    One topic. Extracted claims and structured findings in about 15 minutes. First scan free. How to read Exposure Data.

    One topic · Extracted claims · Structured findings · About 15 minutes

  2. 02

    Exposure Snapshot

    A rapid briefing: notable exposures with severity scores, short explanations, example model output, and cited public sources. Packaged for presentation, including a one-page summary you can share with leadership.

    Up to 3 points of exploration · Headline findings, severity scored · Real-world source attribution · Packaged for presentation · Up to 48 hours

  3. 03

    Basis Report

    The complete package: full findings with evidence excerpts, a prioritized exposure inventory, corroborating model outputs, derivation of how each conclusion was reached, a numbered Sources and Citations registry, glossary, exploitation implications, and access to the underlying raw data. Packaged for presentation. Remediation playbook optional.

    Up to 10 points of exploration · Complete inventory & evidence graph · Raw data access · Packaged for presentation · Cited sources, glossary, next steps · Up to 72 hours

Compare Exposure Data, Snapshot, and Basis Report → · How to read Exposure Data →

Outside-in first, red-teaming only when authorized

Exposure Data, Snapshots, and Basis Reports use public sources only. White-box reachability assessments instrument internal RAG and LLM workflows and always require written authorization. Organizations that need repeatable analysis under their own control can also license the software. See white-box engagements. For the conceptual background, see Sensitive Information Reachability on the SenTeGuard blog.

Research

Related research

Longer conceptual pieces on information reachability and related idea-security work are published on the SenTeGuard blog (company research site, not this product domain).