Quick Answer
Suno’s November 2025 breach exposed 55.3 million user records, but only became public in July 2026 when Have I Been Pwned indexed the data. Exposed fields include email addresses, phone numbers, names, physical addresses and partial card details. Suno never notified users. Check your address on Have I Been Pwned and watch for targeted phishing.
Key Takeaways
- Have I Been Pwned indexed 55,282,226 unique email addresses from the Suno dataset on July 20, 2026
- The breach itself happened in November 2025, an eight-month gap between incident and public disclosure
- Suno says individual notifications were not warranted under applicable privacy laws
- Partial card data cannot be charged on its own, but combined with your name and address it makes impersonation calls very convincing
- The same intrusion took source code that has become evidence in ongoing music industry litigation
What happened in the Suno data breach?
On July 20, 2026, the breach notification service Have I Been Pwned added 55,282,226 unique email addresses to its database, all drawn from a dataset attributed to Suno, one of the most widely used AI music generation platforms.
The intrusion itself did not happen in July. It happened in November 2025. The dataset only began circulating publicly in mid-July 2026, roughly eight months later, after independent outlet 404 Media published a report based on information supplied by the person claiming responsibility.
Suno spokesperson Rachel Racusen did not dispute the number of users affected and confirmed the company experienced a security incident in November 2025, according to Music Business Worldwide. The company has characterized the incident as limited, involving primarily outdated source code no longer in use.
For scale: Suno’s chief executive said in February that more than 100 million people had used the service. If you have ever generated a track on it, even once, even years ago, assume your address is in the file until you have checked otherwise.
What data was actually exposed?
The exposure is uneven, which matters for working out your own risk. The bulk of the dataset is contact information. A much smaller slice is far more sensitive.
| Data type | Scale | Practical risk |
|---|---|---|
| Email addresses | 55.3 million unique | Targeted phishing, credential stuffing on reused passwords |
| Phone numbers | Where supplied at signup | Smishing, voice impersonation, SIM-swap targeting |
| Names and physical addresses | Stripe purchase records only | Identity verification bypass, convincing impersonation |
| Purchase amounts and history | Stripe purchase records only | Highly credible “billing problem” scam calls |
| Partial card details (type, expiry, last four) | Stripe purchase records only | Cannot be charged alone; strong social-engineering proof |
The Stripe portion runs to tens of thousands of records rather than tens of millions. If you never paid Suno anything, you are almost certainly in the email-and-phone tier rather than the full-profile tier.
Published descriptions of the dataset do not list passwords. That is genuinely good news, and it is the single most important detail for most people reading this. It does not eliminate the risk, because roughly 24% of the addresses in this dataset had already appeared in earlier breaches, and a reused password from any one of those is still a live problem.
How did the breach happen?
According to the account the attacker gave 404 Media, entry came through a single compromised employee, not through a flaw in Suno’s own product. The mechanism was Shai-Hulud, a self-replicating supply-chain worm that spread through the npm package ecosystem starting in late 2025.
Shai-Hulud works by harvesting credentials from developer machines and continuous-integration environments: GitHub personal access tokens, cloud API keys, and similar secrets. In an unusual twist, it exfiltrates what it steals to a public repository under the victim’s own GitHub account, which leaves the stolen credentials openly readable by anyone who finds them.
That detail matters for understanding what actually happened here. As supply-chain security firm Socket noted in its analysis, it is not established whether the person who published the Suno data planted the trojanized packages themselves or simply found an infected employee’s exposed credentials sitting in a public repository and walked in. The second scenario fits how the hacker described their own motives, which amounted to opportunism rather than any specific interest in Suno.
Either way, the lesson generalizes badly for the industry: one developer’s laptop was enough to reach source code, the customer list and payment records. This is the same structural weakness behind a growing share of incidents, and it is why agentic AI security risks increasingly centre on credentials and permissions rather than model behaviour.
Why did nobody get notified?
This is the part that has drawn the most criticism, and it is worth stating precisely rather than emotively.
Suno’s position is that the customer information involved was limited enough that individual notifications were not warranted under applicable privacy laws. US breach notification requirements are set state by state, and many are triggered by specific categories such as full payment card numbers, Social Security numbers or account credentials, none of which appear in the published descriptions of this dataset. A company can therefore be entirely within the letter of several state statutes while its users learn about a breach from a third-party service eight months later.
Whether that reading holds up is now a question for regulators rather than for commentators. Several observers have noted that if authorities disagree, the outcome could tighten disclosure expectations across AI platforms generally. As of publication, no enforcement action has been announced.
The practical takeaway for readers is less about the law and more about expectations: you cannot rely on being told. Checking breach databases periodically is not paranoia, it is the only reliable signal available. Our guide on how to check whether your email has been in a data breach walks through the process for any service, not just this one.
How do you check if you were affected?
Before you do anything else, one warning. Breaches of this profile reliably attract a second wave of scams: emails and calls claiming to be from the affected company, from your bank, or from a “breach response” service, all designed to harvest exactly the details the original theft missed. Suno has not notified users directly. Treat any message claiming to be a Suno breach notification as suspect, and never follow a link or call a number supplied inside one.
- Check your address. Go to Have I Been Pwned directly by typing the address into your browser, and enter the email you used to sign up. It will tell you whether that address appears in the Suno dataset and in any earlier breach.
- Check any secondary addresses. If you signed up with a throwaway or work address, check those too. People forget which one they used.
- Change reused passwords first. If the Suno password was used anywhere else, change it on those other accounts before anything else. That is the highest-value action available to you.
- Turn on two-factor authentication on your email account above all others, since email is the reset path for everything else you own.
- Watch your card statement if you ever paid Suno. Partial card details cannot be charged, but the accompanying name, address and purchase history make a fraudulent “we’ve detected a problem with your recent payment” call unusually convincing.
- Report anything that does happen. US readers can file at IdentityTheft.gov, which generates a recovery plan and an official affidavit.
If you get a call that seems to know your recent purchases, hang up and dial the number printed on your card. A caller who can recite your address and the last four digits of your card has proven nothing except that they have read this dataset.
What did the leaked source code reveal?
The breach took more than customer records. It also took internal source code, and that is where the story stops being purely a security matter.
According to 404 Media’s reporting on the leaked material, the code and its accompanying documentation describe a large-scale training-data acquisition pipeline drawing from sources including YouTube Music, Deezer, Genius, Pond5 and others, with scraping reportedly routed through commercial proxy infrastructure. Reported figures from the documentation include tens of thousands of hours from individual sources.
Suno is currently defending copyright infringement claims brought by major labels and coordinated by the RIAA, and argues that training on copyrighted works constitutes fair use. Internal documentation of what was collected and from where is, at minimum, awkward material to have circulating while that litigation is live. Warner Music Group settled separately with Suno in November 2025 and entered a licensing partnership.
These remain allegations tested in court rather than settled findings, and the leaked material has not been independently authenticated in a legal proceeding. But the security lesson stands on its own: a credential compromise did not just cost Suno its customer list, it cost the company control of its own evidentiary position. If you are weighing which generative tools to trust with your data, our roundups of AI music generators and AI video generators are a reasonable place to compare alternatives.
What does this mean for AI platforms generally?
Three things, and none of them are specific to music.
First, consumer AI platforms have accumulated enormous contact databases very quickly, often faster than their security programmes have matured. A service founded in 2023 holding 55 million email addresses is a meaningful target that may not yet be resourced like one.
Second, the supply chain is now the soft entry point. Shai-Hulud’s damage is still surfacing months after its first waves, and Suno is unlikely to be the last organisation to discover it was affected. Any company whose developers install open-source packages is exposed to this class of attack.
Third, disclosure norms in this sector are weaker than most users assume. The gap between “legally required to notify” and “users would obviously want to know” is wide, and it is currently being resolved in favour of the former. Until that changes, treat every AI service signup as a permanent contact-data disclosure. Using a dedicated address for tool signups, and reviewing what you have shared across services, costs very little; our overview of data privacy management tools covers the options for doing this systematically.
FAQ
Was I affected by the Suno data breach?
If you registered a Suno account before November 2025, your email address is likely in the dataset. Enter your address at Have I Been Pwned to confirm. Have I Been Pwned indexed 55,282,226 unique addresses from this breach on July 20, 2026. If you also made a purchase, your name, physical address and partial card details may be included in the smaller Stripe portion.
Did the Suno breach expose passwords?
Published descriptions of the dataset do not list passwords among the exposed fields. The reported data covers email addresses, phone numbers, names, physical addresses, purchase records and partial card details. You should still change any password you reused elsewhere, since around 24% of the addresses in this dataset had already appeared in earlier breaches.
Why did Suno not notify users about the breach?
Suno’s position is that the customer information involved was limited enough that individual notifications were not warranted under applicable privacy laws. US breach notification duties vary by state and are often triggered by categories such as full card numbers or account credentials. The company confirmed a November 2025 security incident but has not issued a direct user notification.
Can someone use the partial card details from the Suno breach?
Not to make a charge. Card type, expiry date and the last four digits are insufficient to complete a transaction on their own. The real risk is social engineering: combined with your name, address and purchase history, those fragments make a fraudulent call claiming to be your bank or the merchant far more believable. Always hang up and call the number on your card.
What did the leaked Suno source code show?
According to 404 Media’s reporting, the leaked code and documentation describe a large-scale training-data pipeline drawing on sources including YouTube Music, Deezer, Genius and Pond5. Suno faces copyright claims from major labels coordinated by the RIAA and argues its training constitutes fair use. These allegations remain contested and untested in court as of publication.