NerdNews

Security personnel are on high alert after discovering 8 compromised TLS certificates with weak private keys, allowing anyone to mint trusted browser and app credentials for any site on the Internet. Browser makers and major apps have blocked the certificates, but updating other parts of the Internet will be a massive and time-consuming task.

Waymo plans to grow its fleet fivefold, from 4,000 vehicles to 20,000 by the end of 2027, using newly secured funding. The company aims to double its fleet by the end of this year and expand its driverless ride-hailing service across the US and globally.

DeepSeek's new open-source inference tool, DeepInfer, could increase pressure on OpenAI and Anthropic by providing a cheaper alternative for businesses. The tool is designed to work with DeepSeek's V4-1 Flash model, which is priced at $0.035 per million input tokens, significantly cheaper than OpenAI's GPT-6 Astra. This move could lead to more businesses adopting Chinese open-source models, posing a challenge to American AI companies.

SpaceX's AI business generated $7.5 billion in revenue, driven by growth in its Super Intelligence business line, including the Cursor coding tool, Grok bot, and neocloud business. The company reported $14.6 billion in third-quarter revenue and $4.4 billion in net income, with AI making up 52% of its income.

Google has rolled out its Gemini Personal Intelligence feature to Android Auto, allowing for more personalized and natural interactions with the AI assistant. The feature enables Gemini to pull information from linked Google apps, providing more specific and relevant responses to user queries. This update improves Gemini's reasoning capabilities and allows for more natural interactions, such as asking for directions to a scheduled appointment or requesting dinner recommendations based on preferences and restrictions.

Large organizations are establishing head of AI security roles to enhance accountability and board-level oversight for AI governance, with 65% of $10bn+ firms having this role, and experience being the biggest driver of compensation differences

TornadoVM 7, a heterogeneous GPU programming framework for Java, enables GPU-native workloads from Java applications. It automatically accelerates Java programs on multi-core CPUs and GPUs, and provides a rich GPU programming ecosystem in Java. The new version brings several architectural changes, including an optimized CUDA backend and support for multi-GPU execution.

OpenAI's GPT-6.1 Sol model has shown significant improvements in migrating legacy code, including languages like COBOL, Fortran, and Visual Basic, with accuracy rates reaching up to 97.9%. This could help companies modernize older software, reducing the need for manual rewriting and minimizing potential errors.

Perplexity's Computer app can now run entirely on a Mac with at least 36GB of RAM, allowing for private and cost-effective AI workflows. Users can choose between Qwen 3.6 and PPLX Pro models, with optional access to over 15 frontier models. The update is available for Perplexity Pro, Max, and Enterprise subscribers.

OpenAI's models can now work with old programming languages like COBOL, Fortran, and Visual Basic, helping banks migrate from outdated systems. The company's GPT-6 Astra model achieved an 87.4% average test pass rate across 20 languages, with 72% for COBOL. This can help banks save time and money by migrating to modern languages like C++, Java, and Python.

A US court has ordered Meta to identify executives who authorized the downloading of pirated books from shadow libraries for AI training, potentially exposing a wider range of supervisors and managers involved in the decision-making process.

A threat actor breached TeamUpdraft's website, potentially compromising UpdraftCentral connection keys, which can send commands to connected websites. The stolen keys are being used to backdoor WordPress sites, with an estimated 10% of sites using UpdraftPlus potentially exposed. Users are recommended to update UpdraftPlus and rotate storage credentials.

Google is upgrading its Gemini AI assistant in vehicles with a personal intelligence feature that connects to Google services like Gmail, Google Calendar, and Maps to provide personalized experiences. The feature will roll out to Android Auto and cars with Google Built-in, allowing users to make complex requests and receive tailored responses.

Eka Robotics has developed a physical AI model that enables robots to pick up objects with ease, valued at $1.45B after a $150M Series A round. The model uses a force-sensitive gripper and a vision-force-action deep learning model, allowing it to adapt to various objects and grasp them without crushing.

Otter AI, a meeting notetaking tool, has raised $320 million from General Catalyst's growth fund. The company plans to use the funds for enterprise growth, go-to-market strategy, international expansion, and product innovation. Otter AI has 40 million users and is approaching new revenue milestones after crossing $100 million in annual recurring revenue last year.

New York City Mayor Zohran Mamdani urged New Yorkers to push back against the rapid rollout of AI, criticizing the federal government's lack of regulation. Mamdani argued that the technology's adoption is not inevitable and that public officials should hold AI companies accountable. He also announced a ban on AI use in elementary and middle schools and introduced a policy requiring landlords to disclose AI-generated listing photos.

Spotify has removed over 50 million AI-generated and spam tracks from user recommendations, prioritizing human-made music. The company updated its recommendation system to identify and remove unwanted tracks, using a combination of listener engagement, upload patterns, and AI detection tools.

Anthropic is having trouble securing enough cloud capacity from major providers, who see the company as a competitor. The company may spend over $750 billion on cloud contracts and is leasing its own data centers due to reliability concerns.

Amulya Vishwanath, founder of Techable Ventures, discusses the future of physical AI, comparing it to the LEGO era where foundational building blocks are being created, and talent is crucial for innovation to scale.

Electric Plant, a startup, has developed an AI model called Bombadil to interpret the electrical signals of plants. The model can detect the needs of plants, such as water and light, and provide real-time feedback to growers. The startup has raised $11.2 million in seed funding and is working to bring its technology to market.