[Survey Results] AI-Related Fines Reach $3.5 Billion — Anthropic and Meta Account for the Majority

A new survey by Surfshark analyzes 10 independent AI-related enforcement actions conducted between 2022 and 2026. Nine of these cases focus on the same issue: "training AI systems using personal, biometric, or copyrighted data without proper consent or legal basis."

Download the graph here

AI-related penalties began in 2022 with three fines targeting Clearview AI. In 2023, Amazon was fined for retaining children's voice recordings used to train Alexa. Regulatory bodies intensified their actions in 2024, with measures taken in four separate cases involving Google, OpenAI, Meta, and Clearview AI. This trend continued into 2025, culminating in a $1.5 billion settlement by Anthropic, the largest in the data analyzed for this survey. Most recently, in 2026, Apple settled for $250 million. This case is the only one where the issue was not the training data itself, but rather the overstatement of AI feature performance.

Dr. Louis Costa, Research Lead at Surfshark (hereinafter referred to as "Costa"), stated:

"The accumulation of massive AI-related fines indicates that regulators are moving beyond mere warnings to imposing significant financial liabilities. While much of the enforcement to date has focused on the use of training data without consent, Apple's $250 million settlement for overstating AI feature performance marks a significant shift in this area.

This may be just the beginning. As seen in the case of OpenAI's fine in Italy being overturned, court decisions still show variability. Nevertheless, the overall trend clearly shows that accountability is catching up with innovation. The AI industry needs to re-evaluate both how it builds AI and how it markets it."

Dr. Louis Costa, Head of Research at Surfshark

Since 2022, regulators and courts have imposed over $3.5 billion (approximately over 560 billion yen) in fines and settlements on seven major technology companies for AI-related violations. Of the 10 cases, Anthropic's $1.5 billion (approximately 240 billion yen) settlement for using pirated books to train AI was the largest single penalty, followed by Meta's $1.4 billion (approximately 224 billion yen) fine for collecting biometric information without consent for facial recognition training. These two cases alone account for 81% of all AI-related fines and settlements in the data analyzed for this survey.

Use of AI Training Data Without Proper Consent or Legal Basis Remains Central to Penalties

Anthropic ($1.5 billion (approx. 240 billion yen)), Google ($291 million (approx. 46.6 billion yen)), and OpenAI ($17 million (approx. 2.72 billion yen) *later overturned*) were criticized for using copyrighted content and personal data to train AI models without proper permission.

Meta ($1.4 billion (approx. 224 billion yen)), Clearview AI (totaling $105 million (approx. 16.8 billion yen) across 4 cases), and Amazon ($25 million (approx. 4 billion yen)) were penalized for collecting biometric information such as facial images and voice data for use in training AI systems.

In contrast, Apple's $250 million (approx. 40 billion yen) settlement was an exception, where the issue was not the acquisition of training data, but the overstatement of AI feature performance.

Not All Fines Are Legally Upheld

In 2024, OpenAI was fined $17 million (approx. 2.72 billion yen) by Italy's data protection authority for using personal data without a legal basis to train ChatGPT and for failing to meet transparency requirements. However, this decision was overturned by an Italian court in 2026.

This case illustrates that the legal framework surrounding AI regulation is still evolving, and what regulators deem a violation may not ultimately be upheld in court.

Meanwhile, Clearview AI was particularly unusual in this analysis, as the company has not paid any of the fines imposed by European regulators.

Clearview AI faced sanctions from data protection authorities in four European countries for the same violation: the unauthorized collection of billions of facial images to build its AI facial recognition database. In 2022, it was fined $23 million (approx. 3.68 billion yen) by authorities in Italy, France, and Greece, and in 2024, it received a $35 million (approx. 5.6 billion yen) fine from the Dutch data protection authority. In total, the fines amount to $105 million (approx. 16.8 billion yen).

However, Clearview AI has not made any payments to date. The company claims it is "not subject to European jurisdiction as it has no physical presence in Europe."

Survey Methodology and Sources

The data for this survey was collected in June 2026, and the findings were published on June 22, 2026.

This survey examined AI-related fines and settlements involving major technology companies between 2022 and 2026. Cases where regulatory bodies or courts took formal legal action against the "training" or "marketing (advertising)" of AI systems were included. The criteria for inclusion were that the company name must be clearly identified in each case, the amount of the penalty or settlement must be measurable, and there must be a direct connection to AI-related activities such as illegal training on copyrighted data or personal data, scraping of biometric data, or misleading claims about AI performance.

The final survey included 10 cases involving 7 companies. For each case, the company name, year of occurrence, penalty amount, and the stated reason for the action were recorded. Penalty amounts were calculated in both Euros and US dollars using an exchange rate of 1 EUR = 0.86 USD. The Japanese yen figures in this article are reference values calculated at an exchange rate of 1 USD = 160 JPY.

When the same company received sanctions from multiple regulatory bodies, each was counted as a separate case. Fines that were later overturned by courts were also included in the survey to provide a comprehensive overview of AI regulation enforcement, with the overturned status clearly indicated.

Details of the materials used in this survey can be found here.

About Surfshark

Surfshark is a cybersecurity company offering an audited VPN, certified antivirus, data leak detection system, private search engine, and online identity generation tools. It has been recognized as a leading VPN service by media outlets such as CNET and TechRadar and was listed in the FT1000: Europe's Fastest Growing Companies.

Its headquarters are in the Netherlands, with offices in Lithuania and Poland. For an overview of Surfshark's business and key initiatives, please refer to its annual summary. For other research projects, please consult the research reports.