AI agents
No deceptive or harmful automation
Agents must not impersonate, manipulate, bypass controls, automate abuse, or make high-impact decisions without oversight.
Acceptable use policy
Clear operating boundaries for agentic workflows, model-driven automation, quantum research tools, embedded Linux, drones, robotics, and public collaboration.
AI agents
Agents must not impersonate, manipulate, bypass controls, automate abuse, or make high-impact decisions without oversight.
Edge systems
Drones, robotics, vehicles, industrial, medical, defence, or public infrastructure deployments require validation and authorization.
Research integrity
Benchmarks, simulations, hardware results, papers, provenance, and reproducibility artifacts must not be fabricated or misrepresented.
01
Policy scope
The AUP covers the website, NowFlow, QFlow Studio, NeuralOS, QANTIS, qmesh, QMANN, managed services, research environments, APIs, integrations, and open-source community channels.
This policy is part of the Terms of Service. A customer agreement, order form, security schedule, or open-source licence may add stricter rules for a specific deployment.
If you use the services through an organization, you are responsible for following both this policy and that organization's acceptable-use, security, compliance, and procurement rules.
02
Prohibited
03
AI safety
AI systems that can act need controls. Users must not configure agents or workflows to deceive, manipulate, harm, or bypass oversight.
Any workflow that affects rights, safety, access, eligibility, employment, credit, health, education, housing, public services, or legal position needs clear authority, human review, logging, testing, and appeal paths.
04
Cyber abuse
05
Privacy
06
Edge and autonomy
NeuralOS, NowFlow, QANTIS, and related tools can support drones, robotics, edge systems, and operational workflows. That does not make every deployment acceptable.
07
Research use
08
Platform integrity
09
Expected controls
Human review
Use approvals, escalation, exception handling, and rollback for actions that can affect people, assets, money, safety, or production systems.
Evidence
Maintain records of prompts, tool calls, data sources, approvals, model versions, deployment packages, and incident handling.
Evaluation
Evaluate outputs, failure modes, bias, latency, safety envelopes, fallback behavior, and model drift before release.
Least privilege
Give agents and users only the credentials, integrations, data scopes, and actions needed for the task.
10
Reporting
Send enough detail for us to identify the issue, affected account, service, timestamp, URL, workflow, repository, device, or integration.
Report policy violations, spam, fraud, impersonation, or misuse.
Report vulnerabilities, credential exposure, suspicious access, or security incidents.
Legal notices, rights requests, or procurement policy questions.
11
Enforcement
12
Current references
Risk-based AI framework, including prohibited practices and obligations phasing in through 2025 and 2026.
UK guidance on fairness, transparency, accountability, and data protection in AI systems.
Voluntary AI risk management framing for governance, mapping, measurement, and management.
13
Updates
We may update this policy as products, AI capabilities, edge deployments, regulatory expectations, or abuse patterns change.
Material updates will be reflected through the last-updated date and, where appropriate, additional customer or workspace notice.
Contact
Include timestamps, workspace IDs, URLs, repository links, workflow names, logs, and screenshots where available.
Policy abuse, impersonation, spam, fraud, or harmful content.
Security vulnerabilities, suspicious access, or incident reports.
Legal notices and formal policy inquiries.