Security

Identity in silicon. Machine identity is anchored in hardware (TPM-backed certificates) and managed through Titan Command Center PKI, with Titan-issued or customer-owned roots. Enrollment uses one-time claim tokens.

A signing gate in front of every skill. Software capabilities load through a signed registry — Titan- or customer-signed in production; unsigned code does not load. The gate has no bypass parameter.

Authority is enforced, not assumed. Every mutating control-plane call passes a role-based permission check and lands in a hash-chained, tamper-evident audit log attributed to a named operator. Remote override of a machine is refused when link latency cannot guarantee a safe handoff.

Encrypted everywhere it moves. TLS on web surfaces; mutually authenticated channels on fleet links. On-device processing keeps raw sensor data local by default; an independent hardware safety layer sits beneath the software stack.

Adversarially exercised. Releases are run through Titan Matrix's failure injection — including sensor spoofing and adversarial-input categories — before they ship. Compliance posture: our enterprise and defense configurations are engineered toward ITAR, FedRAMP, and MIL-STD program requirements; ask [email protected] for current attestation status before relying on any framework.

If you believe you have found a security issue in any Titan AI system — this website, command.titan.ai, our software, or a deployed machine — email [email protected]. Include what you found, where, steps to reproduce, and any proof-of-concept. We accept reports in English, and you may report anonymously.

Our commitments: acknowledgment within 2 business days; an assessment and expected timeline within 7; a fix or mitigation target of 90 days for confirmed issues (faster for anything affecting fielded machines); and credit, if you want it, once the issue is resolved. We will not take legal action against good-faith research conducted under this policy — see safe harbor below.

In scope: titan.ai and its subdomains, the command.titan.ai demo environment, and published Titan software.

Out of scope: denial of service and volumetric testing; social engineering of employees or customers; physical attacks on facilities; any testing against fielded customer machines or production fleets — robots move in the physical world, and testing them without authorization endangers people. If your research would touch a real machine, stop and contact us first; we can arrange a safe environment.

While testing: don't access data that isn't yours beyond the minimum needed to demonstrate the issue; don't degrade the service for others; and don't publicly disclose before we've confirmed a fix or 90 days have passed, whichever comes first — coordinated disclosure is the default, and we'll agree on a date with you.

Research conducted in good faith under this policy is authorized within the meaning of applicable anti-hacking and anti-circumvention laws (including the CFAA and DMCA §1201 in the United States). We will not initiate or support legal action against you for it, and if a third party does, we will make clear your research was authorized. This safe harbor does not extend to testing against fielded machines, customer data exfiltration, or activity outside the rules above.

Reports are triaged by engineering with access limited to those who need it; fixes ship through the same signed, health-gated OTA pipeline as every other release, and the fleet's audit chain records what changed and who approved it. If an incident affects customer data or fielded systems, we notify affected customers without undue delay, with the facts we have and the actions we are taking.

Vulnerabilities: [email protected]. Privacy matters: [email protected]. Everything else: [email protected].