Why we have this policy
AI can help creative businesses spend less time on admin and more time on the work they love. It can also put pressure on the people and industry we exist to serve. We use AI, and we build it into Streamtime, so we owe you a plain account of how we do both.
This policy covers how our team uses AI, how AI works inside Streamtime, and how we keep ourselves accountable.
Our principles
- People first. AI supports human judgment and creativity. It doesn't replace either.
- Privacy and security first. Customer data, employee data and confidential information are protected. When in doubt, it stays out of AI tools.
- Honest about AI. We say when and how we use it. We don't pass AI work off as human work.
- Human review. AI can get things wrong. A person on our team checks our AI-assisted work before it goes outside Streamtime, and takes responsibility for the decision to send it. We have guardrails to catch problems, and we say so when something slips through.
- Experiment with guardrails. We try new tools, within clear safety boundaries.
- Mindful of the planet. We consider AI's environmental footprint when we choose tools and models.
Creative businesses are working out where they stand with AI, and for some that means real disruption. Because of that, we don't use AI to create imagery or video for our own marketing when stock imagery would do the job, or when a customer can provide the animation themselves.
How our team uses AI
- Only approved tools are used with confidential or customer data. Tools outside the approved list are limited to non-sensitive information.
- New tools are assessed for privacy, data handling and training practices, registered, and trialled with non-sensitive data before wider use.
- Customer data, employee records, health information and other sensitive material don't go into AI tools unless specifically approved.
- Higher-risk uses need leadership approval first: customer-facing AI, automated decisions that affect people, processing personal information, and integration with core systems.
- If we use AI to record or summarise a meeting, we tell everyone in it first and stop if anyone objects.
- We build AI skills through peer learning.
AI in our product
- The AI assistant in Streamtime runs on a Claude model from Anthropic, hosted on Amazon Web Services (Amazon Bedrock). It can read your data to answer questions and can propose changes. It asks the person using it to confirm a change before anything is written to your account. What it produces in answer to your instructions is for you to check before you rely on it.
- The Streamtime MCP server lets you connect an AI assistant of your own choosing to your account. The permissions granted at the moment of connection decide what that assistant can do, including whether it can make changes to your data. We don't ask for approval of each individual action, and your chosen assistant has its own confirmation behaviour. Data sent to an outside assistant is handled under that provider's own terms, not ours (clause 14A.8).
- Your data and AI training. Your data isn't passed to third-party model providers to train their models (clauses 14A.3.1 and 14A.4.4). We may use anonymised, aggregated data to maintain and improve the AI features (clause 14A.4.1). We log AI queries and responses to monitor accuracy and fix errors (14A.3.3, 14A.4.2). Our development partners may access Customer Data to build and support the AI features, under confidentiality obligations (14A.3.2).
- Where processing happens and how long it's kept. AI requests are processed in the United States. We keep AI conversations for 30 days after they were last used, then they're deleted. Conversation titles and usage records are kept with your account.
- Your controls. Where we offer the AI opt-out, you can switch off the AI features for your users. The opt-out doesn't stop us storing your data, or using it for our own internal purposes described in clause 14A.4 (such as customer success and product improvement). Changes made through the AI features are recorded in the same audit trail as changes made any other way.
- Our terms. AI features are covered in clause 14A of the Streamtime Licence Agreement, our Privacy Policy (section 4A) and our Data Processing Addendum. The infrastructure and service providers we use are listed at streamtime.net/legal/subprocessors. If this page and the Licence Agreement ever differ, the Licence Agreement governs.
How we keep ourselves accountable
- Our leadership team reviews AI tool use and incidents regularly.
- The Environmental and Social Risk Committee reviews AI risk quarterly, as part of our B Corp recertification work.
- Anyone at Streamtime who sees sensitive data go into an AI tool, harmful AI output, or a customer complaint about AI reports it straight away to the CEO, Head of Operations and Head of Technology.
- We keep a register of the privacy, AI and data laws that apply to us in each country we operate in, and we use it to check AI features before they ship and whenever the law changes.
- We review this policy regularly as the technology, the law and our own experience change.
Human rights and AI
AI can affect people's privacy, fairness and autonomy. We treat these as salient human rights issues for Streamtime. We keep humans in the loop on decisions that affect people, check outputs for bias and inappropriate content, and don't use AI to impersonate a person. These commitments sit alongside our public Human Rights Statement.
Still building
We'd rather tell you what isn't finished than imply it is. We are working on:
- more detailed records of what the AI proposed, what was approved and by whom
- clearer, more specific permissions when you connect an outside AI assistant
- additional usage limits inside the application
Questions or concerns
Write to privacy@streamtime.net, or see how to raise a concern. Policy owner: CEO, Streamtime.










