Responsible AI

AI that serves institutions without undermining the values they protect.

OSIRIS is a powerful AI reasoning system operating in high-stakes institutional environments. These principles define how we build it, how we constrain it, and where we draw lines that cannot be crossed — regardless of what a client requests.

Core Principles

Six principles that govern OSIRIS development and deployment.

Principle I

Human Authority Over Consequential Decisions

OSIRIS produces intelligence. It does not make decisions. Every output — threat classification, escalation forecast, recommended response — is advisory. Human analysts and decision-makers retain full authority over all consequential actions. MIOS is designed to augment human judgment, not replace it.

Principle II

Transparency in AI Reasoning

OSIRIS provides confidence scores, evidence chains, and source attribution with every intelligence output. Analysts can inspect the signals that drove a classification, the network evidence behind an attribution, and the data that informed an escalation forecast. No black-box outputs reach decision-makers.

Principle III

Defensive Intelligence Only

MIOS is built for detection, analysis, and defense — not for offensive information operations. MERDOT does not build, sell, or configure MIOS for the creation of disinformation, synthetic content amplification, or coordinated inauthentic behavior. Any engagement structured around offensive information operations will not be accepted.

Principle IV

No Suppression of Legitimate Discourse

MIOS is designed to detect and analyze coordinated inauthentic behavior — not to suppress legitimate political speech, journalism, or civil society activity. MERDOT will not configure MIOS to target individuals or groups based on political affiliation, religion, ethnicity, or legitimate advocacy activity.

Principle V

Data Minimization and Privacy

MIOS collects and processes only the data necessary for the stated intelligence function. Individual-level personal data is not retained beyond operational necessity. OSIRIS analytics operate at the network and narrative level, not the individual surveillance level. MERDOT complies with all applicable data protection regulations including India's DPDP Act.

Principle VI

Client Accountability

Institutional clients of MIOS accept responsibility for their use of the platform within the bounds of their legal authority and applicable law. MERDOT maintains the right to terminate engagements where MIOS is being used in violation of these principles, applicable law, or in ways that cause demonstrable public harm.

OSIRIS Safety Architecture

How safety is built into the AI engine itself.

OSIRIS includes built-in constraints that operate at the model level — they cannot be disabled by configuration or client instruction. These include:

  • Output confidence thresholds. OSIRIS will not produce high-confidence classifications below a defined evidence threshold. Uncertain outputs are flagged as uncertain, not promoted to false certainty.
  • Attribution constraints. Individual person attribution for information operations requires multi-signal confirmation. OSIRIS will not attribute individual-level blame based on single signals.
  • Demographic profiling blocks. OSIRIS cannot be configured to generate intelligence products that profile individuals by religion, ethnicity, caste, or political affiliation.
  • Offensive content generation blocks. OSIRIS cannot generate disinformation content, synthetic social media posts, or content designed for narrative manipulation.
  • Audit trail requirements. All consequential OSIRIS outputs are logged with full evidence chains. Logs cannot be deleted by client administrators.
Use Policy

What MIOS will and will not be used for.

Permitted Use
  • Detection and analysis of coordinated inauthentic behavior and influence operations
  • Monitoring of information environments for crisis situational awareness
  • Election integrity monitoring and disinformation detection by authorized oversight bodies
  • National security narrative threat monitoring by authorized government agencies
  • Enterprise reputation defense and corporate risk intelligence
  • Academic and policy research on information operations (with appropriate agreements)
Prohibited Use
  • Generating, amplifying, or deploying disinformation or synthetic content
  • Surveillance targeting based on religion, ethnicity, caste, or political beliefs
  • Suppression or targeting of journalists, civil society, or legitimate political opposition
  • Offensive information operations against domestic populations
  • Electoral manipulation on behalf of any political party
  • Any use that violates applicable national or international law

Questions about responsible use of MIOS?

We are committed to transparency about how OSIRIS works and what it will not do. If you have questions about our AI principles, responsible use framework, or specific deployment considerations, contact us directly.