GMO investigates suspected Onamae cloud breach as unproven 'AI attack' claims circulate

What happened

GMO Internet is investigating a suspected breach of Onamae, its cloud desktop service, after customers reported unauthorized activity on trading accounts. According to reporting from Agentic Affairs syndicated on dev.to, as many as 27,000 servers could be affected — a potential scope rather than a confirmed count, with no verified figures yet for how many systems were accessed or what data was taken.

GMO has taken the service offline, reported the matter to police and begun collecting customer devices to examine them for malware. The attackers also issued an extortion demand of roughly 8 billion yen, according to the reporting, which emphasizes that this is the amount demanded — not a loss — and gives no indication of whether GMO paid, negotiated or refused. Material allegedly published by the attackers is still being checked for authenticity.

An especially sensitive target

Onamae supplies virtualized desktops used by financial traders. That makes reports of unauthorized operations on trading accounts more serious than a typical data leak: the possibility is that someone was acting inside live customer accounts, not merely copying files. GMO's decision to shut the service down and involve law enforcement reflects an active, serious incident rather than a routine disclosure.

The Artex connection remains unproven

Some coverage has framed the incident as an "AI attack," pointing to Artex, a Chinese AI-enabled penetration testing tool. There is genuine context for concern: GMO's own cybersecurity unit separately reported evidence of Artex being used against companies in Japan, and CrowdStrike has linked the tool to attacks on South Korean financial institutions.

But no one — neither GMO nor the reporting — has connected Artex to the Onamae breach. That the tool is active in the region does not establish that it was used here, and Agentic Affairs warns that early "AI attack" framing risks pushing AI defense products for an intrusion that has not been diagnosed.

One detail cuts against the AI narrative: attacks associated with Artex have reportedly exploited weak basic controls, including missing multi-factor authentication. In other words, intruders are getting through gaps that routine security hygiene would have closed, rather than defeating advanced defenses. Japan's Financial Services Agency had already urged institutions in May to strengthen vulnerability management and defense in depth.

Why it matters

The incident is serious on its own terms: a suspected compromise of trading infrastructure, reported unauthorized account activity and a multi-billion-yen extortion demand. The risk is that an unproven "AI attack" label distorts the lesson. If investigators find the entry point was missing MFA or weak controls, the practical takeaway for cloud and financial operators is to enforce basic protections — not to rush out and buy AI security tooling. Until GMO or investigators confirm the cause, the 27,000 figure should be read as potential scope, the 8 billion yen as a demand, and the Artex link as unverified.