Alters — Safety Infrastructure for Autonomous Systems

$ cat mission.txt Alters builds safety infrastructure for autonomous systems. We test whether AI agents, autonomous workflows, and machine-to-machine interactions behave safely, ethically, and within their operational boundaries. Our tools evaluate behaviour continuously, adversarially, and with cryptographic proof of what instructions the system was running. Compliance proves controls exist. We prove they work.

$ cat services.json

{
  "safety_evaluation": {
    "description": "Continuous adversarial testing of AI system behaviour",
    "dimensions": ["ethical", "safety", "legal", "bias", "instruction_adherence"],
    "method": "privacy-preserving headless evaluation",
    "output": "Trust Badge — verifiable credential (W3C DID)"
  },
  "identity_infrastructure": {
    "description": "Cryptographic identity for autonomous systems",
    "product": "alters.id",
    "standards": ["W3C DID", "Verifiable Credentials"],
    "features": ["agent registration", "trust scoring", "credential issuance"]
  },
  "provenance": {
    "description": "System prompt certification and auditable memory",
    "product": "Memory Commons",
    "features": ["prompt hash verification", "conversation provenance", "audit trail"]
  },
  "consulting": {
    "description": "Enterprise deployment of AI safety and governance controls",
    "focus": ["regulated industries", "financial services", "government"],
    "contact": "daniel@alters.io"
  }
}

$ cat approach.md How we think about safety Trust is a control outcome, not a communications outcome. We don't certify that your AI controls exist. We prove they work — continuously, adversarially, and with cryptographic proof of what instructions the agent was running. Every evaluation runs privacy-preserving: conversations are scored without human reviewers seeing the content. Only scores persist. Raw data is deleted after evaluation. The ecosystem alters.io — the company (you're here) alters.id — identity, credentials, trust badges /memory-commons — system prompt provenance, auditable memory Open source github.com/altersio/trust-badge

$ echo $CONTACT daniel@alters.io Sydney, Australia $ echo "Founded 2026. Safety engineering for the agentic era." Founded 2026. Safety engineering for the agentic era.