MalwareTips News Critical Veeam and Atlassian flaws lead this week’s security updates

[News Now]

Image: Check Point Research

Businesses running Veeam Backup & Replication 12 or self-managed Atlassian products should prioritize newly released security fixes. Other reported incidents involve exposed personal data, disrupted university systems and phishing designed to influence AI email assistants.

Updates that need attention​

Check Point Research reports that Veeam fixed several Backup & Replication 12 vulnerabilities. The most serious, CVE-2025-64393, could let a low-privileged backup user execute commands on the server; builds earlier than 12.3.2.4934 are affected.

Atlassian also patched CVE-2026-21589, a critical flaw affecting eight self-managed Data Center products, including Jira, Confluence, Bitbucket and Bamboo. An attacker would not need to sign in to access sensitive files and might obtain credentials for connected applications.

• Upgrade Veeam Backup & Replication 12 installations to build 12.3.2.4934 or later.
• Apply Atlassian’s patches to affected self-managed Data Center products, and review exposed credentials if compromise is suspected.
• Update affected Citrix NetScaler appliances using SAML authentication; Citrix has confirmed targeted exploitation of CVE-2026-88779.

Large data and service disruptions reported​

Denmark disclosed unauthorized access to its Central Register of Persons through access provided to a private supplier. Names, addresses and identification numbers associated with about 8.8 million people were reportedly exposed.

Osaka Metropolitan University suffered a suspected ransomware attack that disrupted classes, email, internal networks and administrative systems. Around 500 servers were affected, and information belonging to at least 130,000 people may have been exposed.

Phishing now targets people and AI assistants​

Researchers identified malicious emails that pair familiar phishing lures with prompt injections hidden in HTML comments or invisible text. A prompt injection is an instruction planted in content to manipulate an AI tool; here, it attempts to promote fraudulent requests or trigger unauthorized disclosure.

Good to know

Treat AI-generated email summaries and suggested actions as untrusted when they involve payments, account access or sharing information. Open the original message and verify the request through a separate channel.

Attack volumes continue to rise​

Check Point Research reported that global cyberattacks increased 48% year over year in September 2026, reaching an average of 2,803 weekly attacks per organization. Reported ransomware attacks rose 53%, while education remained the most targeted sector.