# Will the EU AI Act general-purpose AI rules lead to a formal fine against any AI company before 31 December 2028?

> As of 15 November 2026, Prolepsia puts the chance of yes at 75%.

- Forecaster: Prolepsia (https://prolepsia.com)
- Forecast made: 2026-11-15
- Question type: Yes / no
- Resolves: 2028-12-31
- Status: Open
- Forecasts: 1
- Page: https://prolepsia.com/forecasts/calculus-epsilon-lemma/will-the-eu-ai-act-general-purpose-ai-rules-lead-to-a-formal-fine-against-any-ai
- Open in the app: https://app.prolepsia.com/library/b5cc3be3-8718-4265-90fb-8fd81c355f9c

## Forecast

- Yes: 75%
- No: 25%

## How Prolepsia reasons

## TL;DR
The forecast is **75% Yes and 25% No** for at least one qualifying Commission fine adopted between November 15, 2026, and December 30, 2028. Operational enforcement powers, active supervision, and several routes to a monetary penalty support Yes. The decisive distinction is between securing compliance and formally imposing a fine: only the latter counts.

## Context
The EU’s general-purpose AI regime is already in force at the forecast date. GPAI obligations began applying on August 2, 2025, and Commission enforcement powers began on August 2, 2026. Models placed on the market before August 2, 2025, generally have a transitional compliance deadline of August 2, 2027. That distinction applies to models, not whole companies, so established providers are not categorically outside enforcement until the later date. ([Commission GPAI guidance](https://digital-strategy.ec.europa.eu/en/faqs/guidelines-obligations-general-purpose-ai-providers))

The question is narrower than whether Europe penalizes an AI company. It requires a Commission monetary-fine decision under Article 101 connected to GPAI-provider activities. Investigations, warnings, proposed fines, commitments without fines, and penalties exclusively under other laws do not qualify. Adoption by December 30, 2028, Brussels time is what matters; payment and a final, unappealable judgment are unnecessary. A decision disclosed afterward can count if its timely adoption is reliably verified by March 31, 2029, subject to the withdrawal and annulment conditions.

## Evidence
The historical backbone is the Commission’s recent enforcement of adjacent digital laws. These comparisons show that a first monetary decision can fit inside the remaining window. They do not supply a dependable base rate: the reference class is small, selected, and drawn from different legal regimes.

Under the Digital Markets Act, compliance obligations began on March 7, 2024. The Commission opened the relevant investigations on March 25, 2024, and imposed its first Apple and Meta fines on April 23, 2025. That was 394 days from investigation opening to fines, or 412 days from the compliance start. The two companies’ simultaneous decisions are one regime-level timing example, not independent evidence about how quickly a new enforcement programme matures. ([Compliance start](https://digital-markets-act.ec.europa.eu/designated-gatekeepers-must-now-comply-all-obligations-under-digital-markets-act-2024-03-07_en?prefLang=mt); [investigation opening](https://digital-markets-act.ec.europa.eu/commission-opens-non-compliance-investigations-against-alphabet-apple-and-meta-under-digital-markets-2024-03-25_en); [fine decisions](https://ec.europa.eu/commission/presscorner/api/files/document/print/en/ip_25_1085/IP_25_1085_EN.pdf))

The Digital Services Act comparison was slower. X’s relevant compliance obligations began on August 25, 2023; formal proceedings opened on December 18, 2023; and its first fine arrived on December 5, 2025. The interval was 718 days from formal proceedings and 833 days from compliance. Measured from the initial October 12, 2023, information request, it was 785 days. These different starting points matter because the verified GPAI activity remains largely supervisory, rather than a publicly advanced fining case. ([DSA enforcement history](https://digital-strategy.ec.europa.eu/en/policies/list-designated-vlops-and-vloses); [formal proceedings](https://digital-strategy.ec.europa.eu/en/news/commission-opens-formal-proceedings-against-x-under-digital-services-act))

There are 776 elapsed days from the forecast date to the actual qualifying-decision deadline. Legacy models have only 516 elapsed days between their transitional compliance deadline and that endpoint. The inference is straightforward: an early-starting case has time to finish, but a difficult case launched after the legacy deadline faces a tighter calendar. Neither historical comparison makes completion automatic.

The strongest current evidence is that supervision has begun. In its September 1, 2026, briefing, the Commission confirmed information requests involving more than 30 AI companies, covering safety/security and copyright. This is a count of companies contacted, not established violations or independent prospective fines. ([Commission briefing](https://audiovisual.ec.europa.eu/fr/media/video/I-293797))

The Commission’s October 12, 2026, Scientific Panel account confirms that the AI Office had requested and received serious-incident reports and sent information requests about security practices. Further incident-related questions were planned. That is stronger evidence than a legal mandate alone: the regulator is examining actual provider conduct. But receiving an incident report does not establish that the incident, the response, or the reporting breached the Act. ([Scientific Panel report](https://digital-strategy.ec.europa.eu/en/news/special-meeting-ai-scientific-panel))

The legal routes are broader than substantive safety violations. Article 101 covers specified failures involving information, documentation, cooperation, and model access for evaluation, as well as relevant substantive obligations. Its maximum fine is €15 million or 3% of preceding-financial-year worldwide annual turnover, whichever is higher. That is a ceiling, not a minimum; any qualifying monetary fine is enough for this question. ([Article 101](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-101))

This breadth matters. A first case need not establish that a model caused a major harm or resolve the hardest systemic-risk dispute. A sufficiently clear information or access failure provides another route. The Commission’s current enforcement overview confirms these supervisory and enforcement powers. ([Enforcement framework](https://digital-strategy.ec.europa.eu/en/policies/enforcement-ai-act))

Detailed procedures are also adopted. Implementing Regulation 2026/1755 provides for preliminary findings, access safeguards, and at least 21 days for written observations. That response period is not a full investigation timetable. The procedures support progression to a fine but also permit closure without one. ([Procedural regulation](https://eur-lex.europa.eu/eli/reg_impl/2026/1755/oj/eng/pdf))

Technical capacity is developing alongside the legal machinery. Capgemini’s November 5, 2026, announcement identifies a consortium selected to support cyber-risk evaluation tools, compliance methodologies, and on-demand assessments. This is evidence of investment in enforcement capacity. It is a contractor announcement, not proof that assessments have been delivered or that a penalty is planned. ([Consortium announcement](https://www.capgemini.com/wp-content/uploads/2026/11/2026_11_05_Capgemini-selected-to-lead-European-Commission-consortium-for-AI-Safety.pdf))

The counterweight is that successful enforcement can end without punishment. Article 93 allows binding systemic-risk mitigation commitments and a finding that there are no further grounds for action. The Commission’s DMA experience also includes closing a separate Apple investigation after engagement on a compliance solution. Both show why active supervision must not be equated with an eventual fine. ([Article 93](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-93); [DMA announcement](https://ec.europa.eu/commission/presscorner/api/files/document/print/en/ip_25_1085/IP_25_1085_EN.pdf))

## What's non-obvious
The less obvious route to Yes is a procedural or documentary case, rather than a headline finding that a frontier model is unsafe. Article 101 gives the Commission several ways to impose a qualifying fine. That widens the opportunity for a first decision without requiring every difficult scientific and legal issue to be settled. It does not establish that any provider has committed such an infringement. ([Article 101](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-101))

The less obvious route to No is effective supervision. Providers can improve disclosures, supply information, change practices, or accept commitments without receiving a monetary penalty. More supervisory activity therefore supports the forecast, but its significance depends on whether any matter reaches a formal sanction. The number of companies contacted cannot be read as the number of independent chances of a fine: the same enforcement policy, capacity limits, and legal interpretations affect many files together.

## Uncertainties
No verified advanced Article 101 case has been established. A supplementary enforcement log, checked on November 3, 2026, reported no publicly announced formal proceeding or fine. That is a limited finding about the public record, not proof that confidential work is absent. Public preliminary findings or an identifiable formal fining proceeding would give a firmer view of the pipeline’s maturity. ([Enforcement log](https://reglog.io/enforcement))

The legal boundary of particular incidents also remains uncertain. Reporting about laboratory incidents identifies a dispute over pre-market testing and the Act’s scope. Those reports are leads, not adopted infringement findings. A published Commission interpretation applied to an actual case would help distinguish actionable conduct from incidents outside the relevant obligations. ([Reported scope dispute](https://www.euronews.com/2026/10/14/eus-ai-rules-face-first-enforcement-test-over-lab-incidents))

Capacity is another gap. Institutional staffing and technical procurement do not establish the number of dedicated investigators or their likely case throughput. Nor is there a verified delivery schedule that translates outside technical support into completed decisions. More detail on case stages, investigator allocation, and completed evaluations would narrow this uncertainty.

Finally, cooperation, procedural disputes, amendments, and administrative or political restraint can consume the available time. There is no verified instruction barring fines, so none is assumed. The remaining risk is that an operational regime secures corrections—or leaves cases unfinished—without adopting a qualifying monetary decision by the deadline. That is the core of the 25% No forecast.

## Forecast history

| Date | Forecast |
|---|---|
| 2026-11-15 | 75% chance of yes |

## Sources Prolepsia read

- [Guidelines on obligations for General-Purpose AI providers \| Shaping Europe’s digital future](https://digital-strategy.ec.europa.eu/en/faqs/guidelines-obligations-general-purpose-ai-providers)
- [Commission opens non-compliance investigations against Alphabet, Apple and Meta under the Digital Markets Act - Digital Markets Act (DMA)](https://digital-markets-act.ec.europa.eu/commission-opens-non-compliance-investigations-against-alphabet-apple-and-meta-under-digital-markets-2024-03-25_en)
- [European Commission - Press release](https://ec.europa.eu/commission/presscorner/api/files/document/print/en/ip_25_1085/IP_25_1085_EN.pdf)
- [Audiovisual Service - Point de presse de la mi-journée du 01/09/2026](https://audiovisual.ec.europa.eu/fr/media/video/I-293797)
- [Article 101: Fines for providers of general-purpose AI models \| AI Act Service Desk](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-101)
- [Commission Implementing Regulation (EU) 2026/1755 of 20 July 2026 on detailed arrangements for the conduct of certain proceedings by the Commission pursuant to Regulation (EU) 2024/1689 of the European Parliament and of the Council](https://eur-lex.europa.eu/eli/reg_impl/2026/1755/oj/eng/pdf)
- [Press Release Template](https://www.capgemini.com/wp-content/uploads/2026/11/2026_11_05_Capgemini-selected-to-lead-European-Commission-consortium-for-AI-Safety.pdf)
- [EU AI Act enforcement log: has anyone been fined yet?](https://reglog.io/enforcement)
- [How Does the EU AI Office Enforce the AI Act? (with Lucilla Sioli)](https://artificialintelligenceact.substack.com/p/how-does-the-eu-ai-office-enforce)

## The question

This question asks whether the European Union will formally impose at least one monetary fine on a company for violating the general-purpose AI (GPAI) model provisions of the EU Artificial Intelligence Act (Regulation (EU) 2024/1689) between November 15, 2026, and December 30, 2028, inclusive. The question focuses specifically on enforcement of the obligations applicable to providers of general-purpose AI models, rather than enforcement of the AI Act as a whole.

The EU AI Act's obligations for providers of general-purpose AI models became applicable on August 2, 2025. These include requirements concerning technical documentation, information provided to downstream developers, compliance with EU copyright law, and publication of summaries of training content under Article 53. Providers of GPAI models classified as presenting systemic risk face additional evaluation, risk assessment, risk mitigation, incident reporting, and cybersecurity obligations under Articles 51–55.

The European Commission, supported by the European AI Office, acquired enforcement powers over GPAI model providers on August 2, 2026. Under Article 101, the Commission may impose fines of up to €15 million or 3% of a provider's annual worldwide turnover, whichever is higher, for specified infringements. These include violations of relevant obligations, failure to provide required information, and failure to cooperate with certain Commission requests. The rules apply to qualifying providers regardless of whether they are headquartered inside or outside the EU. Models placed on the EU market before August 2, 2025, generally benefit from a transitional compliance deadline of August 2, 2027. Sources: [European Commission — GPAI obligations and enforcement](https://digital-strategy.ec.europa.eu/en/faqs/guidelines-obligations-general-purpose-ai-providers), [EU AI Act Article 101](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-101), and [European Commission — Enforcement framework](https://digital-strategy.ec.europa.eu/en/policies/enforcement-ai-act).

As of November 15, 2026, the GPAI enforcement regime is operational, although enforcement is still at a relatively early stage. The Commission has established supervisory mechanisms, published guidance, and developed the General-Purpose AI Code of Practice to facilitate compliance. The forecast concerns whether these enforcement mechanisms will produce at least one formal monetary penalty against a GPAI model provider before the specified deadline, rather than merely an investigation, warning, or compliance order.

### How it resolves

The question resolves YES if, between November 15, 2026, and December 30, 2028, inclusive, the European Commission formally adopts at least one decision imposing a monetary fine on a company or other commercial legal entity under Article 101 of Regulation (EU) 2024/1689 in connection with its activities as a provider of a general-purpose AI model.

A qualifying fine must be imposed pursuant to the AI Act's enforcement provisions governing GPAI model providers. Qualifying grounds include an infringement of the obligations established by Articles 53–55, failure to comply with a Commission request or measure covered by Article 101, failure to supply required information or documentation, or failure to provide access to a GPAI model for an evaluation where required under the relevant provisions.

The fine must be imposed through a formal administrative decision by the European Commission. A publicly announced enforcement decision counts even if the company subsequently challenges the decision before an EU court, provided the decision has not already been withdrawn or annulled when this question is resolved. The fine does not need to have been paid or collected by the deadline.

The question resolves NO if no qualifying fine has been formally imposed by the end of December 30, 2028, Central European Time (CET). Investigations, preliminary findings, statements of objections, warnings, requests for information, compliance orders, proposed fines, and negotiated commitments without an accompanying formal monetary fine do not qualify.

The primary resolution sources will be official European Commission announcements, Commission enforcement decisions, and publications by the European AI Office concerning GPAI enforcement. The EUR-Lex database and the Official Journal of the European Union may provide supplementary confirmation. Official decisions will take precedence over press reporting where sources disagree.

A decision adopted before the deadline but publicly disclosed afterward will count if its official adoption date falls within the specified period and it can be reliably verified by March 31, 2029. If no qualifying decision can be verified by that date, the question resolves NO.

The question will not be annulled merely because the GPAI enforcement framework is amended, enforcement activity is delayed, or the Commission chooses not to impose fines. Such developments are part of the uncertainty being forecast. Annulment would be appropriate only in exceptional circumstances that make the original outcome fundamentally impossible to determine, rather than simply producing a negative outcome.

### Fine print

The relevant deadline is strictly before December 31, 2028. Therefore, a decision formally adopted on December 30, 2028, qualifies, whereas one adopted on December 31, 2028, does not. The deadline is interpreted using the local time applicable to the European Commission in Brussels.

The question covers companies headquartered anywhere in the world, provided they are subject to the relevant provisions of the EU AI Act. Companies such as OpenAI, Google, Microsoft, Meta, Anthropic, and Mistral AI could qualify if they are acting as providers of GPAI models within the meaning of the regulation. Mentioning these companies does not imply that any has violated the law or is subject to an enforcement proceeding.

Only one qualifying fine is necessary for a YES resolution, regardless of the amount. A fine imposed on a subsidiary or another legal entity within a corporate group also qualifies if that entity is acting as a GPAI model provider.

Fines imposed exclusively under other provisions of the AI Act, such as rules governing prohibited AI practices, high-risk AI systems, or general transparency requirements for AI systems, do not qualify unless the same formal decision also imposes a fine under Article 101 in connection with GPAI model provider obligations.

Similarly, penalties imposed under the General Data Protection Regulation, Digital Services Act, Digital Markets Act, national AI legislation, or competition law do not qualify, even when the underlying conduct involves an AI company or general-purpose AI technology.

Periodic penalty payments intended to compel compliance, where legally distinct from an administrative fine under Article 101, do not qualify on their own. Nor do voluntary settlements involving payments that are not formally characterized as fines under the relevant AI Act provisions.

If a qualifying fine is imposed before the deadline and subsequently reduced, the question still resolves YES, provided the fine is not entirely annulled or withdrawn before resolution. Judicial reversals occurring after the question has been resolved will not retroactively change its outcome.

The question concerns formal enforcement action rather than successful collection of a monetary penalty or a final, unappealable judicial determination of wrongdoing.

---

Prolepsia’s forecasting engine made this forecast: it researched the question, weighed the evidence and wrote the report on this page. A probability is not a promise; this one changes as the evidence does.
