AI Transparency Statement

1. What We Use AI For

Creofy deploys AI across the following core platform functions:

Risk levels are assessed under the EU AI Act (Regulation (EU) 2024/1689) Annex III High-Risk list and Article 50 transparency obligations. None of Creofy's current features fall into the High-Risk category under Annex III or the Prohibited Practices list under Article 5.

2. Which AI Models Power Creofy

Creofy is a deployer of third-party AI models. We do not train or operate our own foundation models. The following providers supply the AI capabilities that power our platform:

Each provider is responsible for their own GPAI (General-Purpose AI) model compliance obligations under Chapter V of the EU AI Act. Creofy's obligations as a deployer are set out in our EU AI Act Statement.

3. What AI Does Not Do on Creofy

To be clear about the scope and limits of AI decision-making on our platform:

• No automated decisions with legal effect: Creofy does not use AI to make decisions that have legal or similarly significant effects on individuals (such as credit scoring, hiring, or access to essential services).
• No emotional inference: We do not use AI systems to infer or classify a user's emotional state from biometric signals.
• No social scoring: We do not score, rank, or classify users based on their social behaviour or personal characteristics for purposes unrelated to the platform.
• No real-time biometric identification in public spaces: We do not deploy real-time remote biometric identification systems.
• No training on your outputs without consent: We do not use your generated content or uploaded personal images to train our own AI models. Third-party providers operate under their own terms — see Section 5.

4. Human Oversight

All AI-generated content on Creofy is subject to human review before publication. We maintain oversight through:

• User control at every step: Users initiate generation jobs, review outputs, and choose whether to publish, edit, or discard them. AI does not publish content autonomously.
• Trait-fidelity audit system: An automated quality check (using GPT-4o vision) scores every generated image against the intended influencer's declared traits. Outputs that fail the minimum threshold are retried or flagged — they are not silently passed to users.
• Content moderation: We rely on provider-level safety classifiers (Fal.ai, Runway, HeyGen, ElevenLabs all implement content filters) and our own Acceptable Use Policy enforcement to prevent harmful outputs.
• No fully autonomous agents: Creofy does not deploy AI agents that take actions on external systems (e.g., auto-posting to social platforms) without an explicit user trigger.
• Feedback mechanisms: Users can report problematic outputs via support@creofy.io. Reports are reviewed by a human and used to improve our prompting and audit systems.

5. Limitations and Known Risks

We are committed to being honest about what AI cannot reliably do and where risks exist:

• Identity drift:AI image and video generation is probabilistic. Generated outputs may not perfectly replicate the intended influencer's appearance across every generation. Our trait-fidelity audit system mitigates but cannot eliminate this risk.
• Bias in generated outputs: AI models trained on existing datasets may reflect biases present in those datasets — including under-representation of certain ethnicities, body types, or ages in generated imagery. We actively test and monitor for this and adjust our prompting strategies where we identify systematic bias.
• Hallucinated content: Text and content suggestions generated by AI may contain inaccuracies, outdated information, or fabricated details. Users should verify factual claims before publishing.
• Misuse potential: Like all generative AI tools, Creofy can potentially be misused to create deceptive or harmful content. Our Acceptable Use Policy and enforcement measures are designed to prevent this, but cannot guarantee zero misuse.
• Third-party model changes: Creofy depends on third-party AI providers. Their model updates, outages, or policy changes may affect output quality or availability in ways outside our control.

6. AI Content Disclosure

Creofy supports disclosure of AI-generated content through the following mechanisms:

• Content metadata: Where technically supported by our generation providers, we embed C2PA (Coalition for Content Provenance and Authenticity) or equivalent metadata in generated outputs to indicate AI origin.
• Platform labelling: Within the Creofy platform, all AI-generated content is labelled with the generation type and model used.
• User responsibility: When distributing content on external platforms, users are responsible for applying any required disclosures under the EU AI Act Article 50 (from 2 August 2026), FTC guidelines, UK ASA rules, and the terms of the relevant platform. See our Acceptable Use Policy for details.
• Interaction disclosure: Creofy does not deploy AI-powered chatbots or virtual assistants that could be mistaken for human agents. Our customer support is handled by humans.

7. Regulatory Compliance

8. How We Monitor and Improve

• We track trait-fidelity audit pass rates, reroll rates, and final failure rates via PostHog, segmented by generation type and influencer characteristics, to detect and address systemic quality issues.
• We review AUP violation reports and enforce action within 2 business days.
• We review our AI provider contracts, terms, and safety documentation at least annually, and when providers release material model updates.
• We monitor regulatory developments in the UK, EU, and US relating to AI and update this statement and our compliance measures accordingly.

9. Feedback and Concerns

If you have concerns about how AI is being used on Creofy — including concerns about bias, harmful outputs, or misuse — please contact us at support@creofy.io (subject: AI Concern). We take all such feedback seriously and use it to inform ongoing improvements.

10. Related Policies

• EU AI Act Statement — detailed analysis of our obligations under the EU AI Act
• Acceptable Use Policy — prohibited content and conduct, including AI disclosure obligations
• Biometric Data Policy — how we handle face geometry and voiceprint data
• Privacy Policy — how we collect, use, and protect personal data
• DMCA / IP Policy — copyright and likeness rights in AI-generated content