Infosecurity

JadePuffer Returns With ENCFORGE: A Ransomware Built to Wipe AI Models, Not Just Data

Published

2 months ago

July 21, 2026

The Return of a First-of-Its-Kind Threat

Late last year, security researchers watched something unprecedented: a ransomware campaign orchestrated entirely by a large language model (LLM). The operator, tracked as JadePuffer, used a Langflow instance as its beachhead and let an AI agent call the shots. Now it’s back — and this time, it brought a weapon tailor-made for the machine learning stack.

On July 20, the Sysdig Threat Research Team (TRT) published a detailed breakdown of JadePuffer’s second act. The group re-entered the same vulnerable Langflow server it compromised in its earlier campaign. Inside, it staged a new payload called ENCFORGE: a UPX-packed Go ransomware binary that targets roughly 180 file extensions spanning the modern ML pipeline.

This is not opportunistic spray-and-pray malware. The targeting is surgical.

ENCFORGE’s Hit List: PyTorch, TensorFlow, HuggingFace, and More

Sysdig’s analysis reveals that ENCFORGE explicitly hunts for the artifacts that data scientists and ML engineers spend weeks — sometimes months — building. The binary’s extension list includes:

PyTorch and TensorFlow checkpoints — the saved state of a model mid-training.
HuggingFace SafeTensors weights — the serialized parameters of popular transformer models.
llama.cpp GGUF quantized models — the compressed, deployable versions of LLMs like Llama and Mistral.
FAISS vector indices — the searchable embeddings that power retrieval-augmented generation (RAG) pipelines.
Apache Parquet and TFRecord training datasets — the formatted data that feeds the training loop.
NumPy arrays — the raw numerical backbone of nearly every scientific Python workload.

The binary also includes a --include command-line flag, letting operators append custom extensions per campaign. The help text inside ENCFORGE actually names LoRA fine-tune adapters and legacy GGML weights as examples. The attackers know the ML stack intimately — or at least, they’ve studied it closely enough to know exactly where it hurts.

The Real Cost: Rebuilding, Not Ransom

Here’s the twist that makes JadePuffer genuinely dangerous. Most ransomware groups demand payment in exchange for a decryption key. But ENCFORGE carries no data-exfiltration capability. There is no leak site. No double-extortion playbook.

The threat is the destruction itself.

Sysdig estimates that rebuilding a production-grade model from scratch — after the last clean backup — costs between $75,000 and $500,000 in cloud GPU time and engineering labor. That gap between the backup snapshot and the attack often represents weeks or months of training runs, hyperparameter tuning, and data curation. If the training data sat on the same host as the encrypted model, recovery is blocked entirely until that data is reconstructed.

ENCFORGE sweeps every model variant on shared storage in a single pass. One encryption run can wipe out an entire team’s work for a quarter.

Container Escape in Under Six Minutes

The entry point is familiar but still effective. JadePuffer again exploited CVE-2025-3248, a missing-authentication flaw in Langflow’s code validation endpoint. The U.S. Cybersecurity and Infrastructure Security Agency (CISA) added this vulnerability to its Known Exploited Vulnerabilities catalog back in May 2025.

Once inside the Langflow container, the AI agent ran through reconnaissance and credential harvesting — the same routine seen in the first campaign. When it discovered a mounted Docker socket, it pivoted to fetch the ransomware payload from its command-and-control (C2) server.

That fetch failed inside the container. So the operator — still working through the Langflow RCE channel — rebuilt the delivery mechanism on the fly.

Over five minutes and 24 seconds, the agent iterated through six Python scripts. It converged on a working pipeline that used the mounted Docker socket to spawn a privileged escape container, copied the locker across the namespace boundary via the host’s procfs, and then ran the encryption pass on the host filesystem — completely outside the original container’s isolation.

ENCFORGE itself uses AES-256-CTR with an RSA-2048 key exchange. It kills processes holding file locks before encrypting, then self-deletes after execution. Clean, precise, and hard to trace.

What This Means for AI Teams

JadePuffer’s return signals a shift in how ransomware groups think about targets. Traditional encryption attacks hit databases, documents, and financial records. But the ML stack — checkpoints, weights, vector indices — is becoming just as valuable, and far harder to replace.

For teams running Langflow, securing Langflow instances against known CVEs is the obvious first step. But the deeper lesson is about defense-in-depth for ML infrastructure. Container escape via a mounted Docker socket is a classic vector, yet it still works because many data science environments prioritize flexibility over isolation.

Organizations should consider AI model backup strategies that separate training data and model artifacts from the compute environment. Immutable storage and air-gapped snapshots can close the gap between the last clean copy and the attack. And monitoring for unusual container spawns or procfs access should be part of any ML-specific detection rule set.

JadePuffer isn’t going away. The group proved it can adapt, rebuild delivery chains in minutes, and target the most expensive assets in the modern AI stack. The question is whether the industry will adapt fast enough to keep up.

Infosecurity

G7 Tells the World to Speed Up the Quantum-Safe Encryption Transition

Published

2 days ago

September 6, 2026

Social Media Admin

Quantum Risk Is No Longer a Distant Worry

For years, the threat of quantum computers breaking today’s encryption felt like a problem for the next generation. The G7 just declared that mindset obsolete.

On September 3, under France’s 2026 G7 Presidency, the French National Cybersecurity Agency (ANSSI) — which chairs the G7 Cybersecurity Working Group — published a new call to action. It pushes governments and private organizations to start the quantum-safe encryption transition now, not later.

The document is blunt: reframe the quantum threat from “a distant future problem” to “a near-term threat that demands action across all sectors, not just critical infrastructure.”

Why the Sudden Urgency?

Quantum computers capable of breaking RSA and ECC — the very backbone of public-key cryptography — aren’t here yet. But the G7 notes that “several recent advances suggest an anticipation” of such machines. The exact timeline is uncertain, which is precisely the problem.

Attackers can already harvest encrypted data today and decrypt it later, once quantum machines mature. That’s the “harvest now, decrypt later” scenario that keeps security experts up at night. Waiting for proof that a working quantum computer exists would be a catastrophic mistake.

What the G7 Wants Organizations to Do

The call to action isn’t just a warning. It lays out a practical roadmap for the PQC migration.

First, identify the systems holding your most critical data. Prioritize those for the transition. Then inventory all cryptographic assets, map dependencies, and build a phased, risk-based plan.

The G7 also has a cost-saving tip: integrate post-quantum cryptography (PQC) into products you’re already buying. Replace systems as part of your standard renewal schedule rather than doing emergency rip-and-replace later. Starting early, the document argues, means lower migration costs overall.

Five Priorities for Governments and Industry

The G7 document outlines five concrete priorities that need attention from policymakers and the private sector:

Raise awareness about quantum threats across all sectors.
Develop national PQC strategies, including building an adequate supply of quantum-safe hardware and software.
Focus R&D on advancing PQC through practical innovation.
Build public-private partnerships between government, industry, and academia to grow domestic expertise.
Integrate PQC into cybersecurity requirements and procurement standards.

The document was signed by the national cybersecurity agencies of all G7 members — Canada, France, Germany, Italy, Japan, the UK, and the US — with support from the EU Commission and the EU Agency for Cybersecurity (ENISA).

ANSSI Is Already Moving the Goalposts

This isn’t ANSSI’s first warning shot. Months earlier, the agency announced it would stop vetting products that lack quantum-safe encryption starting in 2027. By 2030, post-quantum security becomes mandatory in procurement for certain security products in France.

That’s a hard deadline. If you sell security products into the French market, the clock is ticking. The G7 call to action suggests other member states may follow suit with similar requirements.

What This Means for Your Security Roadmap

If you haven’t started planning for the quantum-safe encryption transition, this document is your cue. The conversation has shifted from “if” to “when,” and from “someday” to “now.”

Start by taking inventory. You can’t protect what you don’t know you have. Map your cryptographic dependencies, identify crown-jewel data, and begin conversations with vendors about their PQC roadmaps. Many cybersecurity vendors are already preparing for the migration — make sure yours is one of them.

The quantum threat isn’t science fiction anymore. The G7 just made that official. Will your organization be ready when the deadline hits?

Infosecurity

OpenAI Puts $1 Billion on the Table to Arm Critical Services with AI Defenses

Published

2 days ago

September 6, 2026

Social Media Admin

A Billion-Dollar Bet on the Little Guys

OpenAI has committed a staggering $1 billion to put its cutting-edge AI cybersecurity tools into the hands of those who need them most: the people keeping your lights on and your water running. The announcement, made on September 3, outlines a plan to subsidize access to its Daybreak AI models for essential services across the United States and, eventually, the globe.

It’s a direct response to a grim reality. Small municipalities, rural utilities, and local non-profits are getting hammered by sophisticated cyberattacks, yet they often lack the budget and specialized staff to fight back effectively. They are defending aging infrastructure with outdated tools against adversaries who move at machine speed.

This isn’t charity; it’s a strategic move to level a playing field that has grown dangerously tilted.

What Exactly is Daybreak?

For the uninitiated, Daybreak is OpenAI’s dedicated cybersecurity initiative, first unveiled back in May 2026. It’s not a single product but a suite of capabilities that leverages the company’s frontier large language models (LLMs) alongside its AI-coding assistant, Codex. These tools are designed to be deployed by approved defenders for a wide range of security tasks.

By August, OpenAI had evolved this into a two-tier system: Daybreak Red and Daybreak Blue. Red focuses on offensive security—hunting for vulnerabilities before the bad guys find them. Blue is about defense, helping teams monitor, analyze, and respond to threats in real time.

The New ‘Frontline Defenders’ Program

The new initiative, dubbed Daybreak for Frontline Defenders, is all about integration. OpenAI isn’t just handing out API keys. The program is designed to help critical sectors actually embed these AI models into their existing cybersecurity tools, services, and daily workflows. The goal is to make AI assistance as routine as a firewall update.

Which sectors are first in line? Think water treatment plants, electricity grids, local government networks, non-profits, and banking institutions. The rollout starts in the US, but OpenAI explicitly states it intends to expand to partner countries in the coming weeks.

The potential impact is huge. With Daybreak access, a two-person IT team at a rural water authority could review legacy code for flaws, analyze suspicious network activity, and even develop and test fixes—tasks that would typically require a team of expensive security engineers.

A Pilot with MS-ISAC: Putting Words into Action

Talk is cheap, so OpenAI is pairing the pledge with a concrete pilot. They’ve announced a collaboration with the Multi-State Information Sharing and Analysis Center (MS-ISAC). This pilot will pair Daybreak access with guided training and hands-on assistance for an initial group of public sector and water system defenders.

MS-ISAC is a critical piece of the US cyber defense puzzle. It provides threat intelligence, incident-response support, and real-time information sharing to thousands of public-sector organizations. The plan is to start small, develop a repeatable approach, and then expand the partnership over time. It’s a sensible, methodical start.

The Stark Warning That Preceded the Check

This $1 billion pledge didn’t happen in a vacuum. It landed exactly one week after a coalition of over 100 tech and cybersecurity companies—OpenAI included—published an open letter on August 27. That letter was a blunt instrument, warning of a “narrowing window” to act before AI-enabled attacks escalate to a level that puts critical public services at severe risk.

The message was clear: the same AI that powers defensive tools also supercharges attackers. If we don’t democratize access to frontier AI for defenders, we’re essentially handing the keys to the kingdom to cybercriminals.

OpenAI echoed this sentiment in its announcement, stating that the defender’s window “will not stay open indefinitely.” The opportunity, they argue, is to ensure the advantages of frontier AI extend beyond the largest companies and best-resourced security teams, reaching into the communities and institutions whose security affects millions of people.

Beyond this pledge, OpenAI is also working on what it calls a Defense Factory—an automated approach designed to continuously discover, validate, and fix vulnerabilities. It’s part of a broader push to make AI-driven security proactive rather than reactive.

For anyone tracking the intersection of AI and national security, this is a significant development. The question isn’t whether AI will play a role in defending critical infrastructure—that’s a given. The real question is whether the defenders of that infrastructure will have equal access to the tools. With this billion-dollar bet, OpenAI is trying to make sure they do. For more on how AI is reshaping security, check out our analysis of AI-powered threat detection methods and the growing role of automated vulnerability patching tools.

Infosecurity

US and UK Join Forces to Dismantle Scam Centers Behind Billions in Fraud

Published

3 days ago

September 5, 2026

Social Media Admin

A New Alliance Against Cyber Fraud

The United States and the United Kingdom are pooling resources to shut down the sprawling scam centers that have siphoned billions from victims worldwide. A memorandum of understanding signed Thursday commits both nations to parallel investigations and shared intelligence on the organized crime networks behind these operations, many of which are based in Southeast Asia.

U.S. Attorney Jeanine Ferris Pirro met with senior officials from the U.K.’s National Crime Agency and Crown Prosecutor to formalize the agreement. Pirro stated the objective is to “disable” the Chinese gangs that operate these compounds.

How the Partnership Will Work

The memorandum outlines a framework for both countries to identify overlapping cases and decide which jurisdictions will bring charges. The goal is to prioritize cases that can deliver significant mutual impact.

Officials from both sides had already flagged substantial case overlaps. They are now committed to a joint disruption event with private industry partners, scheduled for early October in London and hosted by the National Crime Agency.

The Scam Center Strike Force Takes the Lead

This initiative is spearheaded by the Scam Center Strike Force, launched last November to coordinate U.S. enforcement against cyber-enabled fraud. The numbers are staggering: the FBI reports that cyber-enabled fraud accounts for nearly 85% of all losses reported to the agency. Americans lost over $12 billion to these scams last year — a figure likely far below reality, as many victims never come forward.

Assistant U.S. Attorney Karen Seifert leads the Strike Force. Testifying before Congress in March, she noted the team includes more than 150 personnel, drawing on prosecutors and agents from the FBI, IRS, and U.S. Postal Inspection Service.

Human Trafficking at the Core

These scam centers are not merely criminal enterprises; they are built on human trafficking. Victims are held in compounds across Myanmar, Cambodia, Laos, and neighboring countries, forced to run investment and romance fraud schemes. Chinese syndicates control the operations, often with the complicity of compromised local officials.

Early Wins and the Road Ahead

The Strike Force has already claimed a major victory. The disruption of Prince Group, a Chinese front company used to launder illicit proceeds, led to sanctions from both U.S. and U.K. agencies. The Justice Department also seized roughly $15 billion in bitcoin tied to the company’s CEO, Chen Zhi.

That seizure sent a clear message. But the problem is vast, and the syndicates are adaptive. The new US-UK partnership signals a recognition that no single nation can tackle this threat alone.

For more on related efforts, see how cyber fraud reporting works and the rise of Southeast Asian scam compounds.