The latest: RIKEN says it confirmed the theft of 430 credentials and found evidence of external connections involving some of them. It also found traces of access to registration information for 4,222 Fugaku users.
What to do: Affected users should remove exposed keys from every system that accepts them, replace the credentials and work with system owners to review access. RIKEN has not said every external system was compromised.
Japan’s Fugaku Supercomputer Was Hacked. The Stolen Keys Could Open Other Systems
Japan’s research institute says attackers stole 430 credentials stored on Fugaku, including keys used to reach systems at other organizations. Its warning to users shows why securing the supercomputer is only part of the response.

Jump to a section
Locking one door does not cancel a key that another door still accepts.
That is the problem facing users of Fugaku, Japan’s supercomputer at the RIKEN Center for Computational Science. After unauthorized access forced the system offline, investigators found that attackers had taken credentials that could be used elsewhere. Some of those credentials were used in external connections, RIKEN said in its November 20 update.
The number is striking: 430 credentials, including private keys, were collected and exfiltrated from some users’ home directories. Of those, RIKEN counted 233 private keys without passphrases and 197 with passphrases. The institute said it had found no other misuse beyond the observed connections. It did not say that all 430 keys opened another system or identify the affected external destinations.
That distinction matters. A stolen credential creates a potential route. Whether it actually works, what it can reach and what happened after a connection require investigation at each destination.
What happened to Fugaku?
According to RIKEN’s initial notice, an unauthorized party gained access between November 10 and 13 through a publicly accessible operational-status dashboard and reached an internal operational network. RIKEN detected suspicious activity on November 12, confirmed unauthorized access the next day and suspended Fugaku’s operations and external access on November 14. Users were notified and running jobs were halted on November 15.
The interruption matters beyond the machine itself. RIKEN describes Fugaku’s research uses as spanning fields including climate simulation, drug discovery and disaster prevention. The institute halted running and queued computing jobs; it has not quantified the resulting research impact.
RIKEN said it replaced administrative keys and invalidated users’ public keys used to log in to Fugaku. Those actions help contain access to Fugaku, but they do not automatically revoke credentials accepted by other organizations. In its first notice, RIKEN said it had no evidence at that point that user research data or files had been exfiltrated. That is not proof that none was taken.
RIKEN has not publicly established the precise technical method used to enter through that dashboard. Its November 20 update described steps including separating public services from administrative networks and rebuilding servers. It did not give a restart date in that update.
The dashboard itself raises a leadership question as well as a technical one: which internal systems can a public-facing service reach? A separate analysis on jonasborchgrevink.com examines the Fugaku status page, suppliers and the decisions behind that exposure.
Why a stolen key can outlive the breach
An SSH key is a pair of cryptographic credentials often used to sign in to a remote computer. A user keeps the private key secret and places the corresponding public key on a system that accepts it. As GitHub’s SSH documentation explains, the pair can authenticate a connection without sending a password with each login.
If an attacker copies the private key, deleting it from the breached machine does not remove the corresponding public key from a different server. That other server may still accept it. The owner must remove or revoke the authorized public key at every destination where it is trusted, then create and install a replacement key pair where access is still needed.
Passphrases can add protection to a private key, but they do not make an exposed key safe to keep using. RIKEN’s instruction applies to users with private keys stored on Fugaku, including those protected by passphrases. GitHub likewise advises removing keys that may have been compromised.
The breach may extend beyond the first organization
Research teams regularly work across institutions and need access to shared computing resources. The risk illustrated here is not that this collaboration is inherently unsafe. It is that a trusted key can be copied into an environment where it is difficult to see everywhere it will be accepted.
For a hypothetical example, a researcher keeps a private key on a shared research system and uses it to reach a second institution’s server. If the first system is breached, shutting it down does not by itself revoke the key on the second server. The second institution needs its own records and response. This example explains the risk; RIKEN has not identified a particular downstream victim in its update.
Hacked.com previously reported on stolen AI service keys. Those are not the same technology as SSH keys, and there is no evidence the incidents are connected. The shared lesson is narrower and useful: credentials can carry trust across organizational boundaries, so containment must reach the services that honor them.
What affected users and organizations should do
- Find and revoke exposed access. Follow RIKEN’s notice. For private SSH keys stored on Fugaku, remove the matching public keys from every external server or service where they are authorized. Revoke or replace other credentials stored there, too. Coordinate with the relevant service owners.
- Create clean replacements. Generate new key pairs in a trusted environment and install the new public keys only where needed. Protect private keys with strong passphrases and restrict permissions. Do not reuse the exposed key.
- Check what the keys could reach. Ask administrators to review authentication and activity logs for the relevant accounts, preserve evidence and investigate unusual access. A log showing a connection does not by itself establish what was accessed or changed.
- Watch for impersonation. Verify incident messages through a known RIKEN contact or its official website. Do not send private keys, passwords or login codes to someone offering help.
Our credential leak response guide explains how to map access and investigate without delaying containment. Organizations dealing with signs of a broader incident can also use our business incident response guide.
What about personal information?
RIKEN said its investigation found traces of access to registration information for 4,222 Fugaku users. The data includes names, email addresses, user IDs and other registration details. It said it had not confirmed leakage of identity-verification documents themselves and had not confirmed outside misuse of the personal information. It is responding on the assumption that the accessed information may have been compromised.
If you receive a notice, confirm it independently and follow the instructions relevant to your account. Be alert for messages using the incident as a pretext to request credentials. Our data breach guide explains practical next steps. People with no connection to Fugaku should not assume they are affected simply because they saw the story.
Where AI fits, and what remains unknown
RIKEN said it used AI to help analyze logs during its investigation. That does not mean AI was used by the attackers. As of its November 20 update, the institute had not publicly identified who was responsible, established the full downstream impact or provided a restart date.
The case is a reminder that the perimeter of a serious incident is not just the machine an attacker entered. It also includes the places that still trust what was taken from it.