Software and API Agreement

This Agreement governs authorized use of Safeera software, iFrameOS, supported Cyber Edition functions, software development kits and APIs supplied by Safeera. It applies when accepted with the relevant software or service order. Hardware ownership, subscription rights and third-party software licenses remain separately defined.

1. License and scope

For the accepted term and configuration, Safeera grants the customer a non-exclusive license to use the supplied software and documented APIs for authorized purposes. The order identifies licensed users or Units, permitted integrations, supported functions, paid features, service dependencies and any transfer restrictions. Purchasing hardware does not transfer Safeera’s source code or patents. It also does not give Safeera ownership of customer data or independently developed applications.

Open-source components remain governed by their licenses; third-party model and software restrictions are disclosed with the relevant component. Nothing here overrides those rights or mandatory law. Do not remove proprietary notices, misuse credentials or copy proprietary software beyond the license. Reverse engineering restrictions apply only to the extent lawful and do not restrict rights granted by an applicable component license.

2. Credentials and permissions

Use separate credentials for human users, agents and integrations. Limit scopes to the authorized function, protect secrets and revoke unused access. Job submission, health telemetry, software administration, financial approval, power management and robot actuation require distinct authority. A general inference key is not permission to change a payout destination, cooling setting or physical safety limit.

An application may access a user’s information or act on a user’s behalf only with the required informed permission. Explain the requested scope, recipients, duration and method of revocation. Do not request broader rights as a condition of an unrelated feature. The application operator is responsible for its own lawful notices and handling.

3. Agents and automated actions

Automation is permitted only within documented, supported functions and delegated authority. The customer must define spending ceilings, approved tools and destinations, and any human approval required for consequential actions. Agents cannot grant themselves authority, evade limits or bind another person to a contract without authorization.

The customer is responsible for its applications and instructions, and Safeera remains responsible for its supplied software and commitments. AI outputs may be incomplete or incorrect. Review outputs before relying on them for financial, clinical, legal or physical decisions. No API response alone verifies that a physical action is safe.

4. Equipment and robotics control

Physical control is authorized only for the equipment, operating area and functions expressly approved for that deployment. The Site and Robotics Safety Rules supplied for the deployment apply. Preserve independent emergency stops, local thermal protection, safe state behavior and human intervention. Loss of internet, authentication, billing access or a model response must not be treated as permission to bypass a safety boundary.

Do not use an integration for unauthorized surveillance, weaponization or unsafe operation near people. An API integration does not make a robot, medical device or assistive product certified or approved. Customer-modified controls require the relevant technical and safety validation before use. Safeera may isolate a hazardous function while preserving a safe local state.

5. Charges and service limits

Paid API calls, software tiers and resource consumption are charged only at accepted rates and units. The order or rate schedule identifies limits, rounding and separately charged services. Exceeding a limit does not authorize a silent upgrade. Customer must obtain approval for increased paid scope; Safeera may reject or limit excess use.

Safeera will supply usage information sufficient to understand charges and will correct verified billing errors. Any minimum commitment, renewal, support response time or availability guarantee must be expressly accepted. The underlying Compute Services Agreement governs purchased compute where incorporated.

6. Data and intellectual property

Use of the APIs grants only the data-processing rights necessary to perform authorized requests. Safeera does not obtain a general training license to private prompts, customer code, files, weights, outputs, voice or camera data. Customer-developed applications and optimizations remain the customer’s subject to existing third-party rights.

Use test data that you are entitled to disclose. Access to production or regulated data requires the applicable processing agreement and approved environment. Diagnostic uploads must be limited and explained; authorization to collect equipment health does not authorize indiscriminate collection of private content. Each party protects the other’s confidential information and limits access to authorized recipients.

7. Security reporting and cooperation

Report suspected credential compromise, an unauthorized command or a vulnerability affecting the service promptly to security@safeera.ai. Include the affected system, time, observed behavior and safe reproduction details, without secrets or unrelated personal data. Do not delay reporting while awaiting a complete investigation. Cooperate with proportionate containment and remediation measures.

Testing requires authorization for the target and method. Do not disrupt availability, access another person’s information or persist access to demonstrate an issue. Reporting does not authorize additional testing or create an automatic reward. Safeera will evaluate reports and coordinate an appropriate response without requiring a reporter to waive protected reporting rights.

8. Updates and compatibility

Safeera will identify material compatibility changes and applicable support or end-of-support dates for the licensed offering. Where practicable, nonurgent breaking API changes will be notified at least 30 days before implementation. A longer committed support period remains binding. Urgent security or safety changes may need shorter notice, with an explanation and reasonable migration assistance.

Updates must not silently expand data permissions, convert a free feature to a paid commitment or materially reduce an accepted fixed-term service. Where a necessary change prevents delivery of the agreed service and no acceptable substitute is agreed, the customer may terminate the affected unprovided portion and receive the applicable refund. Safety-critical updates and safe recovery procedures take priority over unattended changes that could create physical hazards.

9. Suspension, termination and effects

Access may be restricted proportionately for a credible security or safety threat, unlawful use or an unremedied material breach, with notice where feasible. Either party may terminate for a material breach not cured within 15 days of written notice, subject to urgent legal or safety action. The order governs expiration and convenience cancellation of paid commitments.

On termination, stop use of licenses that end and revoke associated credentials. Customer information is returned or deleted under the applicable data terms. Open-source rights, lawful copies and a separately granted perpetual license are not extinguished merely by termination of a cloud account. Essential safety behavior must not be defeated by a license check or billing suspension. Accrued payments and confidentiality duties survive.

10. General and liability

Safeera will provide licensed services with reasonable skill and care. No term promises error-free AI output or eliminates an accepted warranty. The liability terms of the applicable equipment or compute agreement govern claims within its scope. For a separately purchased software service without such terms, each party’s liability is limited, to the extent lawful, to fees paid or payable for it in the preceding 12 months, excluding accrued payments, refunds, fraud, gross negligence, willful misconduct, death or injury, physical damage, confidentiality or data protection breaches and non-waivable liability. Free evaluation software does not authorize unsafe physical deployment.

Material amendments require notice and acceptance where required. California law applies subject to mandatory law, and a separately accepted dispute clause controls its scope; otherwise disputes may be brought in a court with jurisdiction. Notices: legal@safeera.ai or Safeera, Attention: Legal, 101 Jefferson Dr., Menlo Park, CA 94025, United States.