AI Wrote the Ransomware. Seriously.

Cloud security firm Sysdig has published findings on what it describes as the world's first ransomware campaign operated entirely by a large language model (LLM) — an AI system capable of understanding and generating human-like text, here repurposed as an attack engine. According to Sysdig's Threat Research Team, the campaign, named JadePuffer, required no skilled human operator.

The attack began by exploiting CVE-2025-3248, a known flaw in Langflow (an open-source tool used to build AI-powered applications). From there, the AI agent worked autonomously: it harvested credentials, mapped connected systems through lateral movement (moving through a network from one compromised machine to others), installed a persistence mechanism via a cron job (a scheduled task that runs automatically), and ultimately reached a production database server.

What happened next is striking. The agent encrypted 1,342 configuration items inside a Nacos service (a configuration management tool used in cloud environments) and deleted the originals. The encryption key was randomly generated and never saved anywhere. The victim cannot recover their data, even if they pay.

The AI narrated its own actions inside its payloads — which Sysdig notes is actually a new detection opportunity for defenders watching network traffic.

The real warning is the speed. The agent recovered from a failed login and produced a working fix in 31 seconds.

What you should do: Patch internet-facing systems immediately, especially anything running older or unmonitored software. Ensure production databases are not directly reachable from public-facing services. Check that credentials are rotated regularly and not stored in plaintext config files.

The UK's National Crime Agency (NCA) and the Internet Watch Foundation (IWF) have issued a stark warning to parents: AI tools are being used to generate child sexual abuse material from publicly available photos of children, according to Infosecurity Magazine.

The IWF recorded a 26,000% increase in AI-generated videos of child sexual abuse in 2025 — 3,440 cases compared to just 13 the previous year. In one confirmed incident, criminals stole images of schoolchildren from a school website and used AI to generate over 100 abusive images, which were then used to blackmail the school.

Two-thirds of this AI-generated material was classified at the most severe level of abuse.

What you should do: Review the privacy settings on any social media accounts where your children appear. Audit who can view photos you have shared, and consider asking schools and clubs to check what images are publicly visible on their websites. Report concerns to the police or the Child Exploitation and Online Protection command (CEOP) at ceop.police.uk.

What ClamAV is: ClamAV is a free, open-source antivirus scanning engine maintained by Cisco's Talos group, widely used inside email gateways, file upload systems, and security tooling at organisations of all sizes.

What it is: CVE-2026-20215 is a flaw in ClamAV's parser for 7z archives (a compressed file format). When processing a crafted 7z file, the parser mishandles a substream count value in a way that causes it to allocate too little memory, then write data beyond the boundary of that allocation — a classic heap overflow (writing data outside a reserved memory block, like overfilling a labelled box and spilling into the next one). This is one of seven flaws patched in ClamAV releases 1.5.3 and 1.4.5, as reported by Help Net Security.

Who's at risk: Any system running ClamAV that scans untrusted or user-submitted archive files — mail gateways, file upload services, and endpoint security tools are all exposed if unpatched.

CVSS: Not yet scored — treat as High until confirmed. Patch this week.

Root cause: The parser fails to validate the substream count before using it to calculate memory allocation size. A value crafted to be unexpectedly large causes the system to allocate insufficient space, after which the parser writes beyond that space when reading the archive's contents. The developer did not enforce an upper bound on the count before allocation.

Attack vector: An attacker crafts a malformed 7z archive and submits it to any service that scans files with ClamAV — an email attachment, a file upload form, or a shared folder under active monitoring. When ClamAV processes the file, the malformed substream count triggers the overflow. Depending on what sits in adjacent memory, this could crash the scanner or potentially be developed into code execution.

Detection strategies: Monitor for unexpected ClamAV process crashes or restarts. Look for scan jobs that terminate without producing a result on 7z files specifically. Unusual volumes of malformed archive submissions from a single source are worth investigating.

Recommended actions:

1. Upgrade ClamAV to version 1.5.3 (recommended) or 1.4.5 immediately — both are available on the ClamAV GitHub releases page.
2. If immediate patching is not possible, restrict ClamAV from scanning untrusted 7z archives at the gateway level as a temporary measure.
3. Enable logging for scanner crashes and set up alerts if the ClamAV daemon exits unexpectedly.

Most people send hundreds of emails a week without realising the protocol was designed with almost no authentication at all. That gap is one of the most commonly exploited entry points for phishing and brand impersonation attacks, according to a Help Net Security feature on email security strategy.

The practical fix for businesses and individuals running their own domains is DMARC (Domain-based Message Authentication, Reporting and Conformance — a standard that tells receiving mail servers what to do when an email claiming to be from your domain fails authentication checks). Without it, anyone can send email pretending to be you.

Check whether your domain has DMARC configured using MXToolbox's free DMARC lookup — enter your domain and it will tell you in seconds. If it shows no record, ask your email provider or IT contact to set one up. Start with a policy of p=none to monitor, then move to p=reject to block fakes entirely.

France's national cybersecurity agency, ANSSI, has announced it will halt certification of security products that lack quantum-resistant encryption (encryption designed to resist attacks from quantum computers, which can break many current standards) starting in 2027, as reported by Schneier on Security. Government agencies and critical infrastructure operators in France are required to use ANSSI-certified products, making this a practical phase-out deadline. Businesses supplying the French public sector should begin evaluating quantum-safe alternatives now — 2027 is closer than it sounds.