About 100 Korean company Claude accounts reported exposed; chats said untouched
Anthropic told South Korea's privacy regulator that about 100 corporate Claude accounts leaked. The data is contact details, not conversations, and the likely cause is outside software.
The short version
- Anthropic reported the leak of about 100 Korean company accounts to the privacy regulator on the 31st of last month
- The regulator is doing a preliminary review before deciding on a formal investigation
- Exposed data may include account numbers, organization IDs, names and work emails
- An Anthropic researcher says prompts, AI answers and payment data were not accessed
- The access reportedly came through a flaw in outside business software, not Claude itself
Why it matters
The scary headline and the actual evidence differ. What is reported is exposure of business contact details, not of what employees typed into Claude. Even so, names and work emails tied to a known AI vendor are useful to anyone running a convincing phishing campaign, and the case shows that a vendor's suppliers are part of your risk.
What was reported
As kangnamtimes on MSN reported, Anthropic filed a notice with South Korea's Personal Information Protection Commission on the 31st of last month. It said information from about 100 Korean company accounts on Claude had flowed outside the company.
The commission says it is still in a preliminary review. If it opens a formal investigation, it plans to run further checks to establish how large the damage actually is.
How the data seems to have got out
The report says the leak appears to have come from an outside security company getting into Anthropic's internal systems. It cites an earlier Washington Post story from the 26th of last month. That story said a US security startup used a weakness in third-party business software to get in.
So the reported route is a vulnerability in someone else's software that Anthropic uses, not a break into Claude's chat service itself. That detail matters, because it changes what customers should worry about.
What was and was not exposed
Anthropic told some business customers that enterprise account numbers, organization ID numbers, names, email addresses and similar work contact details may have leaked. The company says it cut off the affected systems the day it confirmed the problem. It is working with the software vendor to fix the flaw.
An Anthropic technical researcher, Lydia Hally, posted on X on the 8th that account names, emails and pricing plans were accessed without permission. She said user prompts, AI responses and payment information were not. She added that Anthropic will tighten how it manages and monitors third-party software.
What remains unclear
The regulator has not yet confirmed the scale. The statements about what was not touched come from Anthropic's side so far. Nothing in the report names the affected Korean companies or says whether other countries' customers were involved.
What to watch
- Whether the regulator opens a formal investigation and publishes the true scale
- Whether Anthropic or the software vendor names the flaw and confirms it is fixed
- Whether customers outside South Korea receive similar notices
Try it
If your team uses Claude for work, ask your admin which accounts are on a company plan. Warn staff to be wary of emails that mention their AI subscription, since names and work addresses are the sort of details phishers use.
Sources and links
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by kangnamtimes on MSN · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.