Google stops AI-generated tool exploit as NVIDIA launches Star Elastic

Key Takeaways

Google Thwarts AI-Built Zero-Day Attack Plan

Google researchers found a criminal group planning to use an AI-made tool to break two-factor security. The exploit targeted a popular web system tool and could have bypassed login protections. Google worked with the vendor to fix the flaw before the attackers could launch a mass attack. Experts say the code looked like it was written by an AI because of its clean format and strange comments.

Google Stops Hackers Using AI for Mass Attacks

Google says it likely stopped a hacker group from using AI to find and exploit software flaws. The group planned a large-scale attack to bypass two-factor authentication on many systems. Google worked quickly to fix the issue before the criminals could cause widespread damage. The report notes that hackers from China and North Korea are also using AI tools to find new weaknesses.

Hackers Use AI to Create First 2FA Bypass Tool

Google revealed that criminals used an AI system to build the first known tool that bypasses two-factor authentication. This marks the first time AI was used in the wild to find and weaponize a software flaw. The attack targeted a common open-source tool used for managing web systems. Google fixed the problem before the hackers could use it to steal user accounts.

Google Finds First Zero-Day Exploit Made by AI

Google has identified a zero-day exploit created with artificial intelligence for the first time. The tool was designed to bypass two-factor security on a web-based system administration tool. Google worked with the software maker to patch the flaw before a mass attack could happen. State-sponsored groups from China and North Korea are also using AI to find new security weaknesses.

HP Brings 20 New AI Products to India Market

HP has launched over 20 new products in India to expand its AI-driven technology portfolio. The lineup includes new laptops, workstations, and printing solutions for businesses and students. A major highlight is the HP OmniPad 12, an Android tablet with a detachable keyboard for new users. HP also introduced the EliteBoard G1a, a unique keyboard that contains a full AI computer inside.

HP Unveils Android Tablet and Keyboard PC in India

HP launched more than 20 new products in India, including an Android tablet and a keyboard-based computer. The HP OmniPad 12 is a detachable tablet running Android, aimed at students and first-time PC users. It features a 12-inch screen and runs on Qualcomm processors with up to 18 hours of battery life. The company also released the HP EliteBoard G1a, the world's first AI keyboard PC with computing power built directly into the keys.

Ridgewood Students Advance in National AI Challenge

A class from Ridgewood High School advanced to the national semi-finals of the Mott Million Dollar Challenge. Their project, Reading Reimagined, is an AI platform designed to improve literacy and reading engagement. The team has already won $5,250 and will present their work in Flint, Michigan, in June. They have a chance to win up to $25,000 for their innovative student-driven solution.

AI Fraud Prevention Becomes Standard for Payments

Using AI to stop fraud is now a basic requirement for banks and payment processors. Criminals are using automated tools to attack payment systems at a much larger scale than before. Old security methods that rely on simple rules are failing to catch new types of attacks like deepfakes. Modern AI systems can analyze behavior in real time to block bad transactions while letting good ones through.

Addiction Medicine Offers Lessons on AI Dependence

An addiction specialist compares relying on AI to using substances to relieve discomfort. He notes that people often start using AI to help with writing or thinking, then gradually depend on it for everything. This reliance can weaken natural skills like critical thinking and decision-making over time. The author argues that while AI is useful, we must be careful not to let it replace our own mental abilities.

New AI Models Enable Real-Time Human Conversation

Thinking Machines has released new AI models that allow for natural, real-time collaboration. Unlike older systems that wait for a user to finish speaking, these models can interrupt and respond instantly. They process audio, video, and text simultaneously to mimic human conversation flow. This technology aims to solve the bottleneck where users feel disconnected from AI tools during complex tasks.

Companies Struggle to Manage AI Risks and Nightmares

Many companies are finding it hard to manage the growing risks associated with artificial intelligence. Before generative AI, firms focused on basic ethics policies, but new AI agents require updated strategies. Organizations are now facing challenges with data privacy, bias, and the rapid pace of technological change. Experts suggest that current governance programs need significant updates to handle these new threats effectively.

Police Drones Raise Privacy and AI Concerns

Police departments across the country are expanding drone programs into broader surveillance networks. While drones help with search and rescue, they also collect vast amounts of data on public spaces. Concerns exist about how this footage is stored, shared, and combined with facial recognition technology. Many policies focus on flight rules but fail to address the long-term privacy risks of the data collected.

Eric Allam Discusses Durable AI Agent Strategies

Eric Allam from Trigger.dev explains two main ways to build long-lasting AI agents. He compares the replay method with snapshotting to show how agents can maintain their state over time. Allam highlights the benefits of using Firecracker microVMs for handling stateful computing tasks. His insights help developers create AI systems that can perform complex, multi-step jobs reliably.

NVIDIA Releases Star Elastic for Multiple Model Sizes

NVIDIA AI has launched Star Elastic, a new method to train multiple model sizes in a single run. This technology embeds smaller models inside a larger one, saving storage and compute costs. The release includes 30B, 23B, and 12B reasoning models that can be extracted without extra training. This approach helps developers support different performance needs more efficiently.

Sources

• Google researchers uncover criminal zero-day exploit likely built with AI
• Google says it likely thwarted effort by hacker group to use AI for 'mass exploitation event'
• Hackers Used AI to Develop First Known Zero-Day 2FA Bypass for Mass Exploitation
• Google Detects First AI-Generated Zero-Day Exploit
• HP Expands AI PC and workplace portfolio in India with over 20 new products
• HP Launches Over 20 AI Products in India, Including an Android Tablet and a Keyboard PC
• Ridgewood class heads to national semifinals with Reading Reimagined
• AI-Powered Fraud Prevention Is the New Table Stakes in Payment Processing
• What addiction medicine can teach us about depending on AI
• AI's Collaboration Bottleneck Solved
• What Are Your Company’s AI Nightmares?
• Police drone programs raise questions about use of AI, facial recognition
• Trigger.dev's Eric Allam on Durable AI Agents
• NVIDIA AI Releases Star Elastic: One Checkpoint that Contains 30B, 23B, and 12B Reasoning Models with Zero-Shot Slicing