AI Breakthrough: ‘Machine Unlearning’ Enables Models to Selectively Forget Data

Researchers have made a significant breakthrough in “machine unlearning,” enabling AI models to be trained to selectively forget specific information upon command. This addresses major privacy and ethical concerns in AI by providing a method to remove particular training examples from a neural network without degrading its overall knowledge. The process involves identifying and isolating information pathways within the neural network that correspond to the targeted data and then using gradient-based algorithms to erase them. This capability is crucial for compliance with regulations like GDPR and for enhancing data privacy in AI systems. The technology has wide-ranging implications across various sectors, including healthcare, where it can protect patient confidentiality, and finance, for securing transaction data.

Sources:

Waymo and Lyft Partner to Launch Robotaxi Service in Nashville

Waymo is expanding its robotaxi service to Nashville through a new partnership with Lyft, with autonomous driving set to begin in the coming months and public availability expected in 2026. Riders will be able to book the self-driving vehicles through the Waymo app initially, with the option to use the Lyft app in the future. Under the collaboration, Lyft will handle the management of the robotaxi fleet, including maintenance and cleaning, while Waymo will provide its self-driving technology. This expansion follows Waymo’s successful operations in other cities, such as Phoenix, where it offers autonomous rides to Sky Harbor International Airport. The company also recently received a pilot permit to operate at San Francisco International Airport and began offering fully autonomous rides to San Jose Mineta International Airport.

Sources:

NVIDIA Unveils New Platforms to Accelerate AI-Powered Humanoid Robot Development

NVIDIA has announced new cloud-to-robot computing platforms designed to advance the development of humanoid robots powered by physical AI. The company has released an update to its open, customizable foundation model for humanoid reasoning and skills, called NVIDIA Isaac GR00T N1.5. Additionally, NVIDIA introduced Isaac GR00T-Dreams, a blueprint for generating synthetic motion data to train these robots. Several leading robotics companies, including Agility Robotics, Boston Dynamics, and XPENG Robotics, are adopting the NVIDIA Isaac platform. To further support developers, NVIDIA is also releasing new simulation technologies and open-source datasets.

Sources:

Google Cloud Updates Vertex AI Grounding with Google Maps API

On September 18, 2025, Google Cloud announced updates to the Grounding with Google Maps feature within its Vertex AI platform. The update involves the removal of several fields from the API response, specifically ‘grounding_chunk.maps.text’, ‘grounding_chunk.maps.place_answer_sources.review_snippets.author_attribution’, ‘grounding_chunk.maps.place_answer_sources.flag_content_uri’, and ‘grounding_chunk.maps.place_answer_sources.review_snippets.flag_content_uri’. Additionally, the widget context token will now only be returned when an optional ‘widget_token_enable’ input flag is set by the user.

Sources:

The 8th World Legal Tech and AI Summit, held in Berlin on September 18th and 19th, is focusing on the impact of cutting-edge innovations in legal technology, with a significant emphasis on the applications of Generative AI. The summit brings together legal professionals, in-house counsel, and industry leaders to explore advancements in document automation, case management, and other AI-driven solutions designed to transform legal operations. Key discussions at the event revolve around integrating generative AI into legal research for more rapid and precise outcomes and strategically analyzing the impact of both large and small language models.

Sources:

Tigera Launches Security Solution for AI Workloads on Kubernetes

Tigera has launched a new solution designed to secure artificial intelligence workloads running within Kubernetes clusters. This offering addresses the growing need to protect AI applications as they are increasingly deployed in containerized environments. The announcement was part of a series of industry news updates in the DevOps space.

Sources:

Microsoft Invests Billions in Global AI Infrastructure and New Datacenters

Microsoft has revealed a series of investments in purpose-built datacenters and infrastructure to support the global growth of AI workloads and cloud services. The company introduced its newest and largest US AI datacenter, named Fairwater, in Wisconsin, with several more identical facilities under construction across the United States. Additionally, Microsoft announced plans for a new hyperscale AI datacenter in Narvik, Norway, and a partnership to build the UK’s largest supercomputer in Loughton. These projects represent a multi-billion dollar investment and will be connected to Microsoft’s global cloud network of over 400 datacenters.

Sources:

Critical Kubernetes Vulnerability in C# Client Exposes API Server to MiTM Attacks

A medium-severity vulnerability has been identified in the official Kubernetes C# client that could permit an attacker to intercept and alter sensitive communications. The flaw, which stems from improper certificate validation, could expose applications to Man-in-the-Middle (MiTM) attacks, potentially compromising credentials and other confidential data transmitted to the Kubernetes API server. The vulnerability affects all versions of the client up to and including 17.0.13. The recommended mitigation is to upgrade to version 17.0.14 or newer, which correctly enforces trust chain validation.

Sources:

SEEQC, NQCC, and NVIDIA Collaborate on Interface for Scalable Quantum Error Correction

SEEQC, the UK’s National Quantum Computing Centre (NQCC), and NVIDIA have announced a collaboration to create the first digital interface system that connects quantum processors with supercomputing hardware. This new system is designed to support scalable quantum error correction by combining SEEQC’s digital quantum-classical interface with NVIDIA’s CUDA-Q GPU decoders. The goal is to enable real-time, ultra-low latency error correction, which is a critical step for developing large-scale, fault-tolerant quantum computers. Hosted at the NQCC, this technology is expected to significantly improve data throughput efficiency compared to analog systems and will help advance quantum-enhanced AI and heterogeneous computing.

Sources:

Jülich Supercomputing Centre Deploys NVIDIA DGX Quantum to Advance Hybrid Computing

The Jülich Supercomputing Centre (JSC) in Germany has deployed the NVIDIA DGX Quantum system, a platform that integrates quantum computing with high-performance classical computing. This system combines NVIDIA’s Grace Hopper Superchip with Quantum Machines’ OPX1000 controller, achieving a significant reduction in data-transfer latency between the quantum and classical components. The deployment at JSC, which operates Europe’s fastest supercomputer, JUPITER, aims to accelerate research in areas such as qubit calibration, quantum error correction, and the development of hybrid algorithms. This marks a significant step in making quantum computing resources more accessible within leading supercomputing facilities.

Sources:

Solana Co-Founder Warns Quantum Computing Poses Existential Threat to Bitcoin

Anatoly Yakovenko, the co-founder of Solana, has issued a warning that a significant breakthrough in quantum computing could occur within the next five years, posing a serious threat to Bitcoin’s current encryption. Speaking at the All-In Summit 2025, Yakovenko stated there is a 50% chance of such a breakthrough by 2030 and urged the Bitcoin community to migrate to post-quantum cryptography to protect against this future vulnerability. Bitcoin’s current security relies on the ECDSA algorithm, which could be broken by a sufficiently powerful quantum computer, potentially exposing users’ funds. Yakovenko’s call to action emphasizes the need for the cryptocurrency community to proactively address the security challenges posed by the advancement of quantum computing.

Sources:

Progress Adds AI-Powered Coding Assistance to Telerik and Kendo UI Libraries

Progress has announced the integration of AI-powered coding assistants into its Telerik and Kendo UI libraries. This expansion brings AI assistance to a wider range of UI component libraries, including those for ASP.NET Core. The company had previously introduced AI assistants to its Progress Telerik UI for Blazor and Progress KendoReact products. This move aims to enhance the developer experience by providing intelligent coding suggestions and automation within the development workflow.

Sources: