Lumma Stealer, a notorious infostealer that infected hundreds of thousands of Windows computers last year, has resurfaced with renewed vigor, according to a report released Wednesday. The malware, also known as Lumma Stealer, is employing hard-to-detect attacks to pilfer credentials and sensitive files, marking a significant resurgence in cyber threats.
Lumma first appeared in Russian-speaking cybercrime forums in 2022, utilizing a cloud-based malware-as-a-service model. This allowed for a sprawling infrastructure of domains to host lure sites, offering free cracked software, games, and pirated movies, according to Ars Technica. Law enforcement authorities initially hobbled the infrastructure of Lumma last May, after the malware infected nearly 395,000 Windows computers in just two months.
In other tech news, Chinese AI startup z.ai unveiled its latest large language model, GLM-5, which achieves a record-low hallucination rate, according to VentureBeat. The model, which retains an open-source MIT License, also scored a -1 on the AA-Omniscience Index, representing a 35-point improvement over its predecessor. This places GLM-5 ahead of competitors like Google, OpenAI, and Anthropic in knowledge reliability, according to the report.
Meanwhile, researchers at MIT, the Improbable AI Lab, and ETH Zurich developed a new technique to allow large language models to learn new skills without forgetting past capabilities. This technique, called self-distillation fine-tuning (SDFT), enables models to learn directly from demonstrations and their own experiments, according to VentureBeat. Experiments showed SDFT consistently outperformed traditional supervised fine-tuning.
Additionally, Austrian developer Peter Steinberger's open-source AI assistant OpenClaw saw rapid adoption, prompting the release of NanoClaw, a more secure version. OpenClaw offers a powerful means of autonomously completing work and performing tasks across a user's computer, phone, or business using natural language prompts. NanoClaw, released under an open-source MIT license, addresses security concerns raised by OpenClaw's "permissionless" architecture, according to VentureBeat.
Finally, despite recent speculation, Heroku is not dead, according to a post on Hacker News. The author, a former tech lead at Salesforce Heroku, stated that after speaking with friends still working at the company, Heroku remains operational.
Discussion
AI Experts & Community
Be the first to comment