Cisco Talos has released CAIRN, an open-source toolkit that detects AI-integrated malware using metadata alone, without needing to open or execute files. The tool searches for traces called cognitive artifacts, such as API keys and prompt templates, and works with data from VirusTotal. Findings are organized into an SQLite corpus with YARA rules across three layers, offering a new approach to identifying threats that traditional methods may miss.
Key Takeaways
- Cisco Talos discovered CLOSEDQUORUM, malware that uses multiple LLM judges to automate credential theft across Chrome, Edge, Firefox, and crypto wallets
- CAIRN detects AI-integrated malware using metadata alone, without requiring binary execution
- CLOSEDQUORUM supports DeepSeek, Qwen, Mistral, and Gemini, though no wild deployment has been confirmed
Cisco CAIRN spots AI malware using metadata alone
Cisco Talos released CAIRN, an open-source toolkit that detects AI-integrated malware without opening the file. It searches for traces called cognitive artifacts, such as API keys and prompt templates. CAIRN works using metadata from VirusTotal and does not require binary execution. Findings are organized into an SQLite corpus with YARA rules across three layers.
AI malware runs attacks without human input
Cisco Talos discovered CLOSEDQUORUM, malware that uses multiple LLM judges to automate credential theft. The system votes on actions using a JSON structure and targets Chrome, Edge, Firefox, and crypto wallets. It is built in Go and supports DeepSeek, Qwen, Mistral, and Gemini. Cisco Talos found it using the CAIRN toolkit, though no wild deployment has been confirmed.
Ray Dalio warns AI could fuel a damaging bubble
Ray Dalio said AI could drive major productivity gains but also create an investment bubble. He compared this to the dotcom era when Nasdaq rose 86% in 1999 before falling 77%. He also highlighted growing wealth inequality as a key risk. The WEF found 69% of economists expect AI to boost productivity while 61% doubt job growth from data center spending.
Amazon boosts AI and delivery spending
Amazon is increasing AI infrastructure spending through AWS and new silicon partnerships. It added a US$1.9 billion boost to its 2027 Delivery Service Partner program. The move supports higher driver pay and AI safety tools. Amazon is also deepening ties with Anthropic and Qualcomm.
India's senior tech workers return to class for AI training
IIT Kharagpur and upGrad saw a threefold rise in AI course enrollments among experienced tech workers. The program expanded from four to seven courses in months. Enrollees come from 133 Indian cities and several countries, with a median of 12 years of experience. Bengaluru leads demand, followed by Hyderabad and Pune.
How to Close the Gap Between AI Spending and Real Returns
Many companies invest in AI but focus on low-value projects that do not save real money. To get real value, firms should pick problems that are core to their strategy and profitability. Four key steps can help: define value drivers, focus on market position, prioritize strategic problems, and measure AI impact. These actions help align AI spending with real financial results.
4 Companies Making Money with AI Right Now
NVIDIA, Microsoft, Alphabet, and Palantir are earning real money from AI using different strengths. NVIDIA sells AI hardware and software for data centers and autonomous vehicles. Microsoft offers AI cloud services like Azure Machine Learning for healthcare, finance, and retail. Alphabet uses AI in advertising through Google Ads, while Palantir focuses on AI data orchestration for government and industry.
Harvey Uses GPT-6 Astra to Improve Legal Document Drafting
Harvey is an AI startup that helps law firms and legal teams with complex legal work. Its new use of GPT-6 Astra brings more context into drafting legal documents. This produces better formatted and more complete outputs compared with other models. Lawyers can also set personal preferences, such as numbered lists or color coding, to guide the output. The result lets lawyers spend more time on strategy.
New York and Los Angeles Schools Impose AI Moratoriums
New York City and Los Angeles Unified School District have placed limits on AI use in schools due to growing parent concerns. New York City Mayor Zohran Mamdani said K-8 public schools cannot use AI for at least one year. Only some high school classrooms may use AI pilot platforms, and high schoolers will take two AI critical thinking classes. Los Angeles also imposed an AI moratorium for the 2026-27 school year.
ArticleMiner Builds Knowledge Graphs from Scientific Papers
Scientific papers often hide important data in tables and supplementary files that are hard to interpret. ArticleMiner is a framework that reads papers and extracts structured knowledge using ontology guidance. It builds four modules for fields like drug-discovery chemistry and materials science. The system was tested on 163 papers and showed better results than a same-LLM few-shot baseline in most cases.
Sources
- Cisco CAIRN detects AI malware without opening the binary
- AI malware just removed the human from the attack loop
- 'AI will change the world,' but could also create a devastating bubble: Ray Dalio
- Is AI And Delivery Spending Altering The Investment Case For Amazon (AMZN)?
- India’s senior techies with 12 years of experience are going back to class, and AI is the reason
- Closing the Gap Between AI Investment and Financial Return
- These 4 Companies Are Monetizing AI Today
- Harvey turns legal context into stronger drafts with GPT-6 Astra
- Schools hedge their bets on AI as concerns grow among parents
- ArticleMiner: Ontology-Guided Knowledge Graph Construction from Scientific Publications
Comments
Please log in to post a comment.