Trust & Safety

Trust & Safety

How we handle AI safety, model providers, and protecting you from harmful output.

Last updated: 13 July 2026

We keep a short changelog of substantive changes to this page (in particular changes to our AI-model provider list and to our EU AI Act commitments) so that you can tell whether the positioning below is current.

Who we are

The Service is operated by Semly Pro, a sole proprietorship (eenmanszaak) established in the Netherlands, registered with the KvK (Dutch Chamber of Commerce) under number 99448351, VAT ID NL005387029B31, with registered address Hawaiiweg 41, 1339 NW Almere, Netherlands (trading as “SemlyPro”, “we”, “us”, “our”).

SemlyPro intends to incorporate a private limited company (Semly Pro B.V.). On incorporation, Semly Pro B.V. will assume the operation of the Service and the commitments on this page, and references to “SemlyPro” will be read as references to Semly Pro B.V.

If you need to reach an identified point of contact, our contact channels are listed under Reporting concerns below, and our postal contact is the registered address above.

Our approach

SemlyPro helps businesses create, publish, and track content in the AI-search era. AI content platforms carry real safety and trust risks. This page explains how we work to address them. It describes measures we take in good faith; it does not promise that any measure is fully effective, and it should be read together with our Terms of Service, Acceptable Use Policy, Privacy Policy, and Data Processing Agreement.

Who runs the AI

We use large-language models from third-party providers. Currently:

  • OpenAI Ireland Limited — for a subset of content-generation and analysis features.
  • Anthropic PBC (United States) — for a subset of content-generation, analysis, and MCP (Model Context Protocol) features.

We select model providers based on their published safety practices, their data-handling terms, and their ability to serve our EU, UK and Indian customer base with appropriate transfer safeguards.

Because Anthropic PBC is established in the United States, using its models involves an international transfer of data from the EU/UK to the US. Where we (or our providers) transfer personal data outside the EEA or the UK, we rely on the appropriate safeguards — Standard Contractual Clauses, the UK Addendum/IDTA, and the EU–US Data Privacy Framework (and its UK extension) where the recipient is certified — as described in more detail in our Privacy Policy and Data Processing Agreement.

What we do with your data

We do not train publicly-released AI models on your data. SemlyPro does not use Customer Materials or the Content generated for you to train publicly-released AI models. Our contracts with each AI provider require the same on their side by default. This position may change only by our express written notice to you and, where required by law, with your opt-in consent.

Limited internal use of your data to operate, secure and improve the Service — and limited staff access for troubleshooting and support — does occur, and is described in our Privacy Policy and Data Processing Agreement. To be clear about the boundaries of that access:

SemlyPro personnel and authorised contractors may access Customer Materials and Content only as necessary to (a) provide, maintain and secure the Service; (b) troubleshoot, debug and provide support; © investigate abuse, security or Acceptable-Use issues; and (d) review and improve the Service and its models for SemlyPro’s internal, non-public purposes only. Such access is subject to confidentiality obligations, least-privilege / role-based access controls, and access logging. SemlyPro does not use Customer Materials or Content to train publicly-released AI models.

Data you connect or upload. Where you connect third-party accounts or upload files, that data is treated as your Customer Materials and you remain the controller (or, in India, the Data Fiduciary) for any personal data it contains. This includes, among other sources:

  • Google Ads (via OAuth) — campaign, spend, performance and conversion data;
  • Google Analytics 4 (via OAuth / API) — traffic, audience and event data;
  • Google Search Console (via OAuth / API) — query, impression, click and index data; and
  • Excel / CSV files you upload — arbitrary tabular data you import, which may contain personal data.

Our use of Google Ads, Google Analytics 4 and Google Search Console data complies with the Google API Services User Data Policy, including its Limited Use requirements; our full statement will be published at /legal/google-api-limited-use (to be published).

Full detail is in our Privacy Policy and Data Processing Agreement.

Automated and agent-driven publishing (MCP)

Our MCP integration lets AI assistants (for example, Claude Desktop or Claude Code) work with SemlyPro on your behalf, and some MCP tools are write-capable — they can take actions and publish Content to your connected content-management systems agentically, outside the in-app editor.

Because automated writes can bypass the in-app review flow, we build in safeguards designed to keep a human in control:

  • Write and publish actions are scoped by the permissions and connections you have authorised, and can be limited or revoked by you.
  • Publishing tools are designed to present a preview or require a confirmation step before Content is pushed live, so far as technically feasible.
  • You remain responsible for anything an AI assistant or agent publishes on your behalf through your account, just as you are for Content you publish yourself. Please review our Acceptable Use Policy before enabling agentic publishing.

What we do to prevent harmful outputs

  • Grounded generation. Our content pipeline draws on real, cited source material rather than free-generation. This reduces — but does not eliminate — hallucination.
  • Safety filters. We apply filters at the input, prompt, and output stages that are designed to reduce the generation of prohibited content, in addition to the AI providers’ own safety systems. No filter is fully effective.
  • Prompt-injection defences. We employ measures intended to detect and block many inputs designed to circumvent safety guidelines, but no system can detect or stop every such attempt.
  • Rate limits and publishing controls. We limit rate and volume to reduce misuse risk (and to reduce your risk of search-engine penalisation — see below).
  • Human-in-the-loop by default. Content generated by the Service is designed to be reviewed by a human before publication. Our editor, our audit tools, and our publishing UI encourage review. Where publishing is automated via MCP, the confirmation and permission controls described above are intended to preserve that human oversight.

These are measures we take in good faith; they mitigate risk but do not guarantee that harmful, inaccurate or policy-violating output will never occur.

Where SemlyPro is not a substitute for judgement

Content generated by the Service may be inaccurate, incomplete, biased, or fabricated. It is not fact-checked by SemlyPro. Before publishing:

  • Verify every specific factual claim (names, prices, statistics, quotations, product specifications, legal or medical claims).
  • Ensure the tone and framing fit your brand.
  • Make any AI-assistance disclosure that your applicable law requires (see Our EU AI Act positioning and AI-content transparency in India below).

Full detail is in our Terms of Service, section 8 (AI-generated Content).

Content we will not help you produce

Some categories of content are off-limits, regardless of your subscription tier or use case. These include content that sexually exploits children, promotes terrorism, promotes hate or discrimination, defames identifiable people, or violates applicable law. The full list is in our Acceptable Use Policy.

We also decline to help with:

  • Content in regulated professional sectors (health, medicine, financial services, legal advice, cryptocurrency, gambling) unless a qualified human expert reviews it and applicable sector regulation is followed.
  • Content generated for the primary purpose of manipulating search rankings, as characterised by Google’s own spam and content policies.

Publishing safely

Search engines penalise domains that publish AI-generated content in patterns their spam policies flag as abuse — regardless of how good the individual articles are. To help you stay out of trouble:

  • Publish gradually. Match your domain’s authority and editorial capacity.
  • Review before publishing. Every article, every claim.
  • Diversify your content. Don’t publish clusters of near-duplicate variations.
  • Read the current guidance from the search engines you target.

Full guidance is in our Acceptable Use Policy.

Reporting concerns

Our EU AI Act positioning

SemlyPro operates under the EU AI Act (Regulation (EU) 2024/1689). Here is how we understand our role and what we do to comply. This is our current, good-faith understanding; the detailed value-chain classification is set out in our Terms of Service and Data Processing Agreement, and we will update it as guidance develops.

Classification.

  • To the extent we qualify as such, SemlyPro is a deployer of the general-purpose AI systems supplied by our AI providers (currently OpenAI and Anthropic).
  • To the extent we qualify as such, SemlyPro is a provider of a downstream generative AI system in respect of the content-generation and content-analysis features we place on the market under our own brand. The transparency obligations for generative AI systems arise under Article 50 of the EU AI Act.
  • The Service is not intended for, and is not offered as, an Annex III high-risk AI system. We do not represent SemlyPro as a high-risk AI system, and we prohibit high-risk uses (see below).
  • Our customers are deployers of SemlyPro in their own capacity.

Prohibited practices (Article 5). We prohibit our customers from using the Service for the practices listed in Article 5 of the EU AI Act. See our Acceptable Use Policy, sections 3.7–3.10 (EU AI Act prohibited practices, high-risk uses and related restrictions), for the full list.

High-risk uses (Annex III). We prohibit use of the Service for any Annex III high-risk use case (employment, credit scoring, education, essential services, law enforcement, migration, justice) without our prior written consent.

Article 50 transparency (applicable from 2 August 2026). The transparency obligations for generative AI systems apply from 2 August 2026.

  • We are implementing, ahead of the applicable deadline, machine-readable marking of AI-generated Content in a format detectable as artificially generated or manipulated, using solutions that are effective, interoperable, robust and reliable insofar as technically feasible. Under the EU AI Act “Omnibus” transitional arrangements (agreed May 2026), generative AI systems already on the market before 2 August 2026 have until 2 December 2026 to meet the machine-readable marking requirement of Article 50(2); systems placed on the market on or after 2 August 2026 must comply from that date.
  • Where our customers publish deep fakes, or text intended to inform the public on matters of public interest, they are required to make the disclosure required by Article 50(3) and 50(4). We aim to surface tools to make this disclosure straightforward.
  • Our implementation approach and technical documentation of content marking is published at /ai-transparency.

Serious-incident reporting (Article 73, from 2 August 2026). Where and to the extent this obligation applies to us, we will report serious incidents involving the Service to the competent market-surveillance authority within the statutory timeframes (15 days as standard; 2 days for widespread infringement; 10 days where a death is involved). If you are a customer, the obligation to notify us of a serious incident within 24 hours of becoming aware of it is set out in your Terms of Service and, where applicable, your Data Processing Agreement or Master Services Agreement — this page describes that duty but does not itself create it.

Voluntary Code of Practice. We monitor and, where appropriate, intend to align with the EU AI Office’s Code of Practice on Article 50 as it develops, to the extent it applies to us.

AI incident and abuse reporting. If you see the Service being used in ways that may violate EU AI Act obligations, contact anil@semlypro.com (and, for serious incidents, anil@semlypro.com).

AI-content transparency in India

India is a focus market for SemlyPro and the Digital Personal Data Protection Act, 2023 applies to our offering of services to users in India.

AI-content labelling. India is introducing labelling requirements for “synthetically generated information” through proposed 2025 amendments to the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021. As drafted, tools that create or modify such content must apply a prominent, visible label covering at least ~10% of the visual display area (or the first ~10% of audio) and embed permanent, unique metadata or identifiers. These rules remain in draft as of mid-2026, and their final form may change. Our machine-readable marking (described above) is intended to help satisfy the metadata limb; where you publish AI-generated Content to audiences in India, you should apply the required visible on-screen label to meet the visible-label limb. We will confirm and update this guidance when the rules are finalised.

Grievance Officer. For Data Principal requests, consumer complaints and intermediary grievances relating to Indian users, please contact our Grievance Officer at anil@semlypro.com. Details of the notice and consent basis, Data Principal rights, children’s-data protections, cross-border transfers, and breach notification to the Data Protection Board of India are set out in our India Privacy & Grievance Addendum.

United Kingdom

UK users are covered by the UK GDPR and applicable UK guidance. The UK does not currently have a statutory AI-content-marking regime equivalent to the EU AI Act (its approach as of 2026 is principles- and regulator-led), but we monitor UK AI-regulation developments and follow ICO guidance on AI and automated processing. As a non-UK operator that offers services to individuals in the UK, we are addressing UK-establishment matters — including appointment of a UK Article 27 representative and ICO registration and payment of the data-protection fee where required — as described in our Privacy Policy. Our assessment of any Online Safety Act 2023 duties is described there as well; our core service is not a user-to-user service, but we keep this under review for any public-facing surfaces.

Transparency

As we grow, and as the EU AI Act, Digital Services Act, India’s DPDP framework and related regimes phase in obligations that apply to us, we intend to publish additional transparency information here, where and to the extent required by law, including:

  • transparency reports as required by law;
  • our AI-model provider list and any material updates;
  • serious-incident statistics (aggregated); and
  • content-moderation action statistics (aggregated).

Each of these will be published as and when the underlying capability and any applicable legal obligation are in place.

Last updated 13 July 2026 · Operated by Semly Pro (eenmanszaak), KvK 99448351, Hawaiiweg 41, 1339 NW Almere, Netherlands · Questions about this document: anil@semlypro.com

← All legal documents