How We Use AI

The Underlying Models

Polpo 8 does not operate its own AI models from scratch. We use large language models (LLMs) and specialist AI systems provided by OneFirewall.ai — a platform built specifically for enterprise deployment with security, sovereignty, and compliance at its core.

We chose OneFirewall.ai because enterprise AI should not mean "your confidential data goes somewhere you can't audit." Their models are designed to be deployed in controlled, secure environments that meet the requirements organizations like financial services firms, cybersecurity companies, and regulated industries actually have.

The Orchestration Layer

The "central brain" in Polpo 8 is our orchestration system — the logic layer that takes a mission, breaks it down, routes subtasks to the appropriate arms, manages context between them, synthesizes outputs, and manages quality control. This is where our engineering expertise lives.

Think of it this way: the AI models are the engines. Polpo 8 is the aircraft that puts those engines in the right configuration to actually fly somewhere specific.

What AI Does in Polpo 8

Within the Polpo 8 system, AI is used for the following categories of work:

• Generation: Creating text, code, structured data, and other content based on instructions, context, and templates provided by or derived from client requirements.
• Analysis: Processing and summarizing documents, data, and research to extract structured insights. The AI identifies patterns and key information; a human (or a configured review agent) validates before action is taken.
• Classification and routing: Determining what type of task is being requested and which arm or combination of arms should handle it.
• Synthesis: Combining outputs from multiple arms into cohesive deliverables — for example, combining research from the Intelligence Layer with copy from the Content Engine into a finished strategy document.
• Decision support: Generating options, recommendations, and risk assessments for human decision-makers. Polpo 8 informs decisions; it does not replace them on matters of strategic consequence.
• Automation execution: Triggering pre-defined actions in response to conditions or events, based on rules and parameters set by the client.

What AI Does Not Do in Polpo 8

We think it's equally important to be clear about limits:

• AI in Polpo 8 does not make consequential decisions without human configuration and oversight. Clients define the parameters; AI operates within them.
• We do not use AI to generate content that could deceive end users about its AI-originated nature without disclosure.
• We do not use AI to make decisions about individuals in ways that carry legal or significant personal consequences without explicit human review.
• We do not train our models on client data without explicit, documented consent. Client data is used to generate outputs for those clients — not to improve models used for other clients.

Human Oversight

Every Polpo 8 deployment includes configuration of oversight thresholds — points at which human review is required before an output is published, sent, or acted upon. The level of automation is calibrated to the risk level of the action.

High-stakes outputs (customer-facing communications, published content, automated financial decisions) always have a human checkpoint unless the client has explicitly configured and accepted autonomous operation for that task category.

We believe "AI-operated" doesn't mean "human-absent." It means the humans are focused on the decisions that require human judgment, rather than the execution that doesn't.

Data Privacy and Security

Client data processed by Polpo 8 is handled under strict data processing agreements that comply with the General Data Protection Regulation (GDPR) and applicable EU data protection law. Full details are in our Privacy Policy.

We do not share client data across clients. We do not use client data to train AI models without explicit consent. We maintain separation between client environments.

The AI infrastructure provided by OneFirewall.ai offers data residency options for clients with specific sovereignty requirements. If this applies to your organisation, mention it when you contact us.

AI-Generated Content Disclosure

Polpo 8 produces AI-generated content as part of its service. It is the responsibility of clients to disclose AI-generated content to their end users where required by law, platform terms, or professional ethics. We provide documentation to support this where required.

We do not believe in hiding the fact that AI is involved. Clients who are deploying AI-generated outputs to audiences should be transparent about it — both because it's becoming legally required in more jurisdictions and because, frankly, their audiences probably already assume it anyway.

Ongoing Development

The AI landscape changes rapidly. We update our model partners, safety practices, and orchestration logic on a continuous basis. When significant changes are made that affect how client data is handled or how outputs are generated, we notify affected clients.

If you have specific requirements, concerns, or questions about how AI is used in Polpo 8, we'd rather answer them directly. Get in touch.