McKesson's Cloud Accounts and a $55 Million Demand (08/31/2026)

Read the McKesson numbers carefully, because two different figures are circulating and they measure different things. McKesson has confirmed a breach of several cloud-hosted accounts and data exfiltration. The rest is a claim from the group, which has itself said it does not know how many individuals are ultimately affected.

Top 5 Critical Security Alerts

1. Phishing Reached McKesson’s Cloud Tenants

McKesson confirmed that attackers breached several of its cloud-hosted accounts earlier in the week and exfiltrated data, with the affected material relating to oncology, multispecialty, and medical-surgical units, and warned customers of intermittent service degradation. ShinyHunters claims it took millions of rows from Snowflake and Salesforce environments including names, addresses, Social Security numbers, diagnoses, medications, allergies, patient notes, and employee home addresses, and has demanded $55 million. The group says it is unsure how many individuals are ultimately affected. TechCrunch verified a small subset of the samples against public records. Access came through phishing and social engineering against employees. TechCrunch

Operator Note: A 284 million figure is circulating from the group’s own claim, and it counts rows rather than people, which is the same distinction that took the Carhartt claim from a headline number to 12.9 million confirmed. Wait for the notification count before briefing anybody. The operational lesson is elsewhere anyway: nothing was wrong with Snowflake or Salesforce, and an employee was talked into granting access to both. The Record

2. Malware Hidden in PNG Files With a Custom Reverse Tunnel

A campaign is concealing its payload inside PNG image files and dropping a custom reverse tunnel onto victim machines. The image carries the code past controls that inspect file type before content. The Register

Operator Note: This is the fourth reverse tunnel we have carried in five days, after TerminalFix, the router implants, and the agent escape. The pattern is worth naming for your leadership: attackers have converged on outbound connections because inbound is the only direction most organizations actually filter. An image file is also the most permissive object type on your network, since blocking PNGs is not an option anybody will accept.

3. North Korean Job Fraud Moves Into Healthcare and Sales

The employment fraud operation that has placed operatives in IT roles has expanded into healthcare and sales positions. The tradecraft is unchanged and only the target roles have widened. The Hacker News

Operator Note: The IT-worker version of this was containable because technical interviews are hard to fake and technical teams knew to look. Healthcare and sales hiring runs through managers who have never been briefed on it, and a sales role reaches customer data on day one without anybody granting production access. Ask your talent team whether identity verification differs by department, because the answer is almost certainly yes and the reasoning was never security.

4. ValleyRAT Hides in Signed Adware People Add to Exclusion Lists

The backdoor is arriving inside signed adware, which users then add to their antivirus exclusion lists because the product keeps flagging it. The signature is what buys the initial benefit of the doubt. The Hacker News

Operator Note: The exclusion list is doing the work here, and it is a list almost nobody audits. Somebody adds an entry to stop a nuisance alert, the entry is permanent, and the exclusion outlives both the nuisance and the person. Pull yours and read it. Kaspersky’s writeup has the technical detail. Securelist

5. Aurora Ransomware Operators Are Using Cursor AI

The group used the Cursor AI coding assistant during attacks against ten targets. Researchers documented the tooling in use rather than any novel payload. The Hacker News

Operator Note: Ten targets is a small number and that is the point, since it is the first sample rather than the trend. What it tells you is that developer tooling is now part of the intrusion kit, which means your own developer tooling deserves the same scrutiny you give a remote access product.

Additional Security Alerts

Threat Intelligence

• Microsoft has issued its own warning on TerminalFix reverse tunnels: The campaign we carried Saturday now has vendor guidance behind it. BleepingComputer
• Anthropic is cracking down on hijacked accounts being used to mine AI tokens: Follow-up to yesterday’s session theft item, with the stolen sessions now being resold as compute. The Register
• Schneier asks whether somebody is hacking Defense Department refrigerators: Worth reading for the underlying point about what counts as a connected asset. Schneier on Security

Security Standards & Frameworks

• Dark Reading argues that model rules are not security controls: A model instructed not to do something is not the same as a system that cannot. Dark Reading

The Axe Report is a daily briefing from Grab The Axe. Need help assessing your organization’s security posture? Take our free Human Attack Surface Score assessment.