NerdNews

Google has granted the US Department of Defense access to its AI for classified networks, following Anthropic's refusal to do so due to concerns over domestic mass surveillance and autonomous weapons. This deal allows the Pentagon to use Google's AI for all lawful purposes, with some provisions in place to prevent misuse.

Snapchat has launched 'AI Sponsored Snaps', allowing users to interact with brands' AI agents directly. This new format enables personalized, AI-powered interactions and is expected to drive engagement and purchases. With nearly one billion monthly active users, brands can leverage this feature to reach their target audience.

Researchers found that LLMs leave statistical fingerprints in generated passwords. They built a model to detect these fingerprints and found 28,000 LLM-generated passwords in the wild, mostly from Anthropic, Qwen, and Google. This raises concerns about password security and the use of LLMs as password generators.

Elon Musk and Sam Altman are in a trial that will determine OpenAI's future. Musk claims OpenAI has abandoned its nonprofit mission, while Altman says the lawsuit is a baseless attempt to derail a competitor. The outcome could radically change the AI landscape.

OpenAI's top models are now available on Amazon Web Services' Bedrock managed inference and agent platform, providing an alternative avenue for accessing OpenAI's models without exposing data to OpenAI's APIs. This collaboration allows enterprises to build AI-augmented tools using OpenAI's models while addressing security, data privacy, and sovereignty concerns.

Google Cloud introduced Agents CLI to simplify the development lifecycle of AI agents. The tool integrates with coding agents, provides a unified interface to Google Cloud services, and automates infrastructure provisioning and release workflows.

GitHub has introduced a new approach to improving deployment safety by leveraging eBPF, enabling the company to detect and prevent hidden circular dependencies that could block recovery during outages. The technique allows GitHub to monitor and selectively restrict network behavior of deployment processes at the kernel level.

CodeGuardian is a Model Context Protocol server that extends AI assistants with specialized tools for automated analysis and vulnerability detection, providing AI-powered remediation and actual code fixes

Over 3,000 software developers gathered at AI Dev 26 x SF to discuss the future of software development in the AI era. Experts predict that AI will take over many aspects of software development, with some suggesting that small teams of generalists overseeing AI agents will be the way forward. The conference featured presentations from industry leaders, including Andrew Ng, who argued that AI agents should write all of the code, rather than just a portion of it.

IBM's AI coding partner Bob uses a mix of LLMs and open-source models to automate and augment the full software development lifecycle, delivering 45% productivity gains. It integrates with IBM Watsonx Code Assistant for Z and offers advanced features for enterprise-scale mainframe applications.

Autodesk's Flow Studio update introduces AI Rigging and Neural Layer, automating character rigging and reducing rendering costs. This simplifies the animation process, allowing creators to focus on performance and refinement. The tool aims to keep artists in control while streamlining the workflow.

Amazon launched a new AI-powered feature that allows users to ask questions about products and receive conversational audio responses. The 'Join the chat' feature aims to save customers time by providing key product details without requiring them to scroll through lengthy descriptions or reviews.

YouTube is testing an AI-powered search feature called 'Ask YouTube' that provides step-by-step results with a mix of text and video. The feature is currently available to Premium subscribers in the US and allows users to ask follow-up questions.

OpenChoreo 1.0 is an open-source internal developer platform built on Kubernetes, featuring a programmable control plane, AI agents, GitOps, and observability. It provides a complete foundation for running workloads on Kubernetes without requiring custom builds. The project has been accepted into the Cloud Native Computing Foundation (CNCF) Sandbox and has 785 contributors across 240 organizations.

Tenstorrent's Galaxy Blackhole AI servers feature 32 Blackhole accelerators, 1TB of GDDR6, and 23 petaFLOPS of dense FP8 performance, with a price tag of $110,000. The system can be scaled up to 32 nodes and is positioned for large language models and video generation.

Meta has signed an agreement with Overview Energy to receive up to 1 gigawatt of power from spacecraft that collect solar power in space and beam it to solar farms on Earth as infrared light, increasing the return on investment from building solar farms and reducing reliance on fossil fuels.

OpenAI's trial against Elon Musk has begun, but the company is facing more pressing issues, including missing its internal user and revenue goals. This has led to concerns about the company's financial stability and the potential impact on the entire AI industry. OpenAI's CFO, Sarah Friar, has expressed worries about revenue growth, and the company's board has questioned CEO Sam Altman's efforts to secure more computing power despite the business slowdown.

A subgroup of the Lazarus Group, known as BlueNoroff, has conducted a large-scale cyber theft campaign targeting over 100 cryptocurrency organizations across more than 20 countries. The spear-phishing campaign involved multiple social engineering techniques, including impersonating prominent figures, typosquatted Zoom and Microsoft Teams meeting links, and fake Calendly calendar invites. The attackers used AI-generated images to create fake meeting content and deployed a ClickFix-style clipboard injection attack.

The US Supreme Court is considering a case that could redefine digital privacy rights, focusing on the use of 'geofence' search warrants that allow law enforcement to collect location data from tech companies like Google. The case, Chatrie v. United States, centers on a Virginia man convicted of a 2019 bank robbery, where police used a geofence warrant to collect location data from Google. The court's decision could have broader implications for Americans' privacy and the use of geofence warrants by law enforcement.

Neurable, a brain-computer interface (BCI) startup, is looking to license its 'mind-reading' technology to consumer wearables. The company's non-invasive technology uses EEG sensors and AI to analyze brain activity, providing insights into cognitive performance. Neurable has already partnered with companies like HP Inc.'s HyperX and iMotions, and is now seeking to expand its reach across various industries, including health, athletics, and gaming.