AI & Automation Terms
Valkiria uses AI to draft campaigns, creatives, reports and recommendations, and can apply some changes to your ad accounts. AI can be wrong, so you must review what it produces. By default, nothing that creates new spend, launches, raises budgets or deletes anything happens without a person's approval.
1. Scope
These AI & Automation Terms form part of the Terms of Service at /terms and apply to all AI-assisted and automated features of Valkiria, whether used in the app, through messenger bots, the public API or MCP clients. If they conflict with the Terms on AI or automation, these terms prevail.
2. What the AI features do
Valkiria's AI features include:
- VillReach, the AI assistant that answers questions about your account, explains data and prepares actions;
- recommendations for campaigns, budgets, bids, keywords and audiences;
- drafts of campaigns, ad texts, landing and client pages, posts and SEO content;
- generated images, banners, video and voice;
- analyses and reports, including website and market research;
- the AI support chat.
Features vary by plan and may be limited by monthly AI usage allowances.
3. AI outputs can be wrong — you must review them
AI outputs are generated automatically from statistical models and from the data available to them. They can be inaccurate, incomplete, outdated, inconsistent or unsuitable for your purpose, and similar requests may produce different results. Reports may contain estimates, which the app marks where possible.
Before you publish, launch or rely on any AI output, you must check it for accuracy, legality, compliance with platform policies and suitability for your business. AI outputs are not legal, tax, financial or medical advice.
Valkiria does not guarantee any advertising results, such as reach, clicks, conversions, revenue or return on ad spend, and does not guarantee that ads will be approved by any platform.
4. How changes to your ad accounts are approved
Every change that Valkiria's automation proposes for a connected ad account goes through a write policy that decides whether the change is blocked, waits for a person's approval, or is applied automatically.
4.1 Always require a person's approval, at every autonomy level: new spend (such as creating or launching campaigns), deletions, and changes that cannot be reversed.
4.2 Always blocked: all changes while the emergency stop is active or writes are switched off for the account; all automated changes at the "observe" level; live changes on ad accounts where live writes are not enabled, except creating paused campaigns or changes made by a person; and changes beyond the limits of the Starter plan.
4.3 Changes approved by a person are applied, unless they exceed a campaign budget cap or client daily budget limit that the account owner has set; those caps cannot be overridden by an approval.
4.4 Drafts created by generation, publishing, import and quality checks are created as validate-only, draft or paused. Connected AI assistants using the "propose" permission can only create paused drafts and proposals and cannot activate campaigns or increase spend.
5. Autonomy levels
You choose the autonomy level for your account or for each client. The levels are:
- Observe: Valkiria analyses and recommends but makes no changes.
- Manual: every change waits for a person's approval.
- Safe autopilot (the default): reversible changes that reduce spend, such as pausing ads, lowering budgets or bids, or adding negative keywords, are applied automatically. Everything else, including any increase in spend and resuming paused items, waits for approval.
- Auto (opt-in): reversible changes that reduce or keep spend, such as pausing, lowering a budget or bid, or adding negative keywords, are applied automatically if they stay within the account's limits (by default: at most 25% change per step, at most 20 automated changes per hour, reversible changes only). Any increase in spend — a higher budget or bid, resuming paused items — as well as new spend and deletions always waits for a person's approval, at every level. Changes outside the limits also wait for approval.
The account owner can adjust these limits and set budget caps in the automation settings. Changes to limits apply to future decisions only.
You are responsible for choosing an autonomy level and limits that suit your business, for monitoring your accounts and for approvals given by your users. Automated reductions or pauses can affect delivery and results. You can switch to Observe or Manual, or use the emergency stop, at any time.
6. AI providers
Valkiria routes AI requests to models running on GPU infrastructure used for Valkiria first. When that capacity is unavailable or saturated, requests are handled by external AI providers, which may include Alibaba Cloud (Qwen), DeepSeek, Zhipu (z.ai), OpenAI, Anthropic, Groq, Google (Gemini) and Together AI, and for media generation Kling, Higgsfield, OpenAI and ElevenLabs. The current list is at /legal/subprocessors.
Requests sent to AI providers contain the content needed for the task, such as prompts, website, campaign and ad content. Do not include special categories of personal data or unnecessary personal data in prompts.
Own keys (BYOK): on plans that include it, you can connect your own API keys for supported providers. Requests made with your key are processed by that provider under your contract with it, and usage is billed to you by the provider. You are responsible for complying with that provider's terms. We store your keys encrypted and use them only to perform your requests.
7. Use of your data
We use your content and data to provide AI features to you. We do not sell your data, and we do not use your content to train general-purpose AI models offered to third parties.
Valkiria may use your feedback on outputs and usage statistics to improve the quality of its own prompts, rules and support knowledge base. External AI providers process requests under their own API terms; their retention and use of data, including any use for improving their models, is governed by those terms.
8. Rights in generated content
As between you and Valkiria, you own the outputs generated for you, to the extent they are protected by law, and you may use them for your business. Valkiria does not claim ownership of your outputs. AI outputs may not be protected by copyright in some jurisdictions, and similar outputs may be generated for other users.
You are responsible for checking that outputs do not infringe third-party rights, such as copyright, trademarks, personality rights or the rights of persons shown, before you use them, and for obtaining any needed permissions. This is especially important where you ask for content in the style of, or referring to, a real person, brand or work.
9. Prohibited AI uses
In addition to the Acceptable Use Policy at /legal/acceptable-use, you must not use the AI features to:
- create deepfakes or content that impersonates a real person without their consent, or misleading content presented as real;
- generate content that is illegal, sexual content involving minors, or content that promotes violence or hatred;
- produce disinformation, fake reviews or manipulative content, including content that exploits vulnerabilities of specific groups;
- profile or make decisions about individuals that produce legal or similarly significant effects, or engage in practices prohibited by Article 5 of the EU AI Act (Regulation (EU) 2024/1689);
- extract the underlying prompts, models or training data, or circumvent safety filters and usage limits;
- generate malware or content intended for cyber attacks.
10. Transparency and labelling
Valkiria tells users when they are interacting with an AI system, such as VillReach and the AI support chat. If you deploy a Valkiria chatbot or messenger assistant to your own customers, you must make it clear to them that they are interacting with an automated AI system, unless this is obvious from the context.
Where the law requires it, including under Article 50 of the EU AI Act once applicable, you must disclose that images, audio or video are artificially generated or manipulated (in particular deepfakes), and that published text informing the public on matters of public interest was AI-generated unless it was editorially reviewed. You must also follow the AI-content labelling rules of each ad platform and social network.
You must not remove or alter any labels, watermarks or metadata that indicate AI generation.
11. Responsibility and liability
You remain responsible for your advertising, your content and decisions taken on the basis of AI outputs, including changes you approve or allow under your chosen autonomy level.
Valkiria's liability is governed by the Terms at /terms. Nothing in these terms limits liability that cannot be limited by law, including for intent, gross negligence, death or personal injury, or mandatory consumer rights. If an automated change is applied by Valkiria in breach of the rules described in section 4 or 5, this is a defect of the service for which we are responsible under the Terms.