The through-line in this beat's recent logging is not that breaches keep happening, but that the value has migrated from personal data to credentials that unlock AI systems. A regulatory fine in Sweden, a phishing campaign dressed as an Anthropic offer, and stolen AI logins across tens of thousands of corporate domains are the same story from three angles: attackers are going after the keys to AI accounts, and the organisations holding those keys are not defending them well enough.
A Record Fine for Weak Vendor Security
Sweden's data privacy regulator, IMY, has imposed a $183,000 (SEK 1.8 million) fine on IT systems provider Miljödata, according to BleepingComputer, over inadequate security measures leading to a breach in August 2025 that affected 2.2 million people. The figures are modest next to the population touched, and that mismatch is the point. A single supplier's omissions travelled through every organisation that depended on it, which is the classic pattern of a systemic data breach rather than a single compromised target.
For US technology companies, the Miljödata outcome is a reminder that supplier-side failures are now priced and punished. A fine of that size will not deter a large vendor, but it establishes a documented fault line. US firms selling into Sweden or handling EU residents' data operate under similar expectations of adequate security, and a vendor breach that cascades to millions of records is exactly the kind of exposure that regulators on both sides of the Atlantic have been signalling they will pursue.
Phishing That Targets the Login, Not the Wallet
Malwarebytes reported, and CNET covered, a new phishing attack that promises Claude Max but steals Google credentials instead. The offer appears as a website promotion from Anthropic, and the campaign is harder to detect than most. This is credential theft dressed in the branding of an AI provider. The attacker does not need to breach Anthropic; it needs only to convince a user to hand over the identity that unlocks their connected accounts.
The choice of Google credentials as the prize connects directly to the third story. In corporate environments, Google sign-in is frequently the front door to AI tools and the mailbox, documents and sessions behind them. A user pursuing a premium AI subscription is doing something entirely ordinary. That ordinariness is what makes the lure effective, and it is what makes this a data breach story rather than a consumer-fraud curiosity. Stolen Google credentials are an entry point, not an end state.
80,000 Domains and a Market for Stolen AI Logins
Infostealer logs exposed AI account credentials and sessions tied to more than 80,000 corporate domains, according to SOCRadar research reported by BleepingComputer. The same reporting describes risks running from stolen conversations to LLMjacking, and a growing market for stolen AI logins. The scale is the argument: this is not opportunistic theft of a few accounts but an inventory of corporate AI access being collected and traded.
The connection to shadow AI is direct. Credentials can only be stolen from systems that exist, and many of these AI accounts were in use without security teams tracking them. Once an AI login is captured along with a session, the attacker inherits the assistant's context: documents passed in, questions asked, drafts produced, and any integrated data the session can reach. The category the SOCRadar reporting puts at the centre, LLMjacking, turns stolen access into a resale problem, where someone else's AI capacity is consumed at the victim organisation's expense.
