AI Disclosure & Limitation of Liability

Last updated: April 12, 2026Version 1.0StableState LLC

1. AI Disclosure

The IronFist AI platform deploys autonomous AI agents powered by large language models (LLMs) provided by Anthropic. Your agent uses these models to interpret instructions, execute multi-step tasks, generate text, and make decisions about tool usage.

AI agents are not infallible. They can produce incorrect, incomplete, or inappropriate outputs. They can misinterpret instructions. They can take unintended actions when given ambiguous procedures.

2. Accuracy & Reliability

StableState does not guarantee the accuracy, completeness, or appropriateness of any output generated by your agent. You are responsible for reviewing agent actions and outputs, especially in high-stakes domains including (but not limited to) financial decisions, legal communications, medical information, public-facing content, and personnel decisions.

Your agent's behavior depends on: the quality of instructions and procedures you provide, the capabilities and limitations of the underlying LLM, the accuracy and availability of data from your connected integrations, and network conditions and third-party API availability.

3. Approval Gates

By default, high-risk tool categories require your explicit approval before execution. These categories include: external messaging (sending emails, Slack messages, social media posts), financial operations (Stripe actions, payment processing), data deletion (removing records, files, or entries), and browser write actions (form filling, clicking on external sites).

Your agent will request approval via your communication channel before executing any action in a gated category. You review the proposed action and approve or reject it.

4. Approval Gate Opt-Outs

You may opt out of approval gates for specific categories via the dashboard. Each opt-out requires a separate, per-category acknowledgment that is stored with a timestamp in your tenant record.

When you opt out of an approval gate for a category, you acknowledge that: your agent will perform actions in that category without requiring your approval, you accept full responsibility for the consequences of ungated actions, and StableState is not responsible for the content, recipients, or impact of actions taken by your agent after the gate is removed.

You may re-enable approval gates for any category at any time via the dashboard.

5. Limitation of Liability

StableState's total liability for any single claim arising from or related to the Platform is limited to the greater of: $1,000 USD (minimum floor), or the total subscription fees paid by you in the 12 months preceding the incident giving rise to the claim.

This cap applies per incident, not in aggregate.

6. Exclusions from Liability

StableState is not liable for: actions taken by your agent using your API keys and credentials, costs incurred on your third-party accounts (Anthropic, Google, Microsoft, etc.), inaccurate or harmful LLM outputs, upstream provider outages (Anthropic, Azure, Google, Microsoft, Browserbase), data loss resulting from your failure to export data during the post-cancellation retention window, consequences of approval gate opt-outs that you have explicitly acknowledged, or actions taken by your agent that violate the Acceptable Use Policy.

7. Acknowledgment

By accepting this document, you acknowledge that: your agent is powered by AI that can produce incorrect outputs, you are responsible for reviewing agent actions in high-stakes domains, approval gates exist to protect you and opting out shifts responsibility to you, and the liability limitations described above are fair and reasonable given the nature of the service.

Questions about this disclosure? Contact [email protected].

Questions about this document?

Visit https://www.stablestateit.com for general inquiries and our contact form.