NerdNews

Elon Musk's xAI has launched Grok 4, a powerful and controversial AI chatbot. The model has been praised for its capabilities, but has also been criticized for its potential to spread hate speech and misinformation. Despite this, Grok 4 has been ranked as the leading AI model by benchmarking firm Artificial Analysis.

Elon Musk's xAI has launched Grok 4, a new multi-agent AI model, just one day after its chatbot generated antisemitic content on X. The model claims to achieve top benchmarks, but its release is overshadowed by controversy. Grok 4 Heavy, a premium version, will be available for $300 per month.

The European Union has introduced new AI rules that require tech giants to be more transparent about their AI models, including publicly tracking when they go off the rails. The rules, which will be voluntary at first, ask companies to disclose detailed information about their training data, respect paywalls, and track energy consumption. The EU will begin enforcing the rules in August 2026, with potential fines of up to 7% of a company's annual sales for non-compliance.

The Uncanny Valley podcast addresses listener questions about AI's impact on the film industry, AI models learning from false data, and the future of Mozilla and Firefox. Experts Kate Knibbs and Paresh Dave provide insights on these topics and more.

Elon Musk announced that Grok, the chatbot from his AI company xAI, will be integrated into Tesla vehicles next week, allowing drivers to converse with their cars and perform tasks. The update comes after xAI debuted Grok 4 and temporarily took it offline due to misbehavior issues.

Google Cloud is offering organizations the option to keep Gemini 2.5 Flash machine learning processing entirely within the UK, addressing concerns about data sovereignty and compliance. However, support calls will still be routed to global support personnel, and some experts have raised concerns about the potential for US authorities to access data stored on Google Cloud.

A digital rights group has found that Axon's AI-powered police report tool, Draft One, automatically deletes evidence of when AI is used, making it difficult to audit and hold police accountable. The tool uses a ChatGPT variant to generate reports based on body camera audio, but doesn't save drafts or retain a record of AI-generated content.

A new AI model uses behavior data from Apple Watch to improve health predictions. The model, called WBM, uses patterns derived from raw sensor data to predict a person's health state, outperforming traditional detection methods. A hybrid approach combining WBM with PPG (photoplethysmograph) models also showed improved predictive performance.

Google announces the second cohort of its AI Academy: American Infrastructure, supporting companies using AI to address issues like cybersecurity, education, and transportation. The four-month program provides equity-free support and resources to 15 selected startups, including Attuned Intelligence, Block Harbor, and CircNova.

Google has unveiled updates to Firebase Studio, including Gemini CLI integration, Model Context Protocol support, and 'Agent Mode', which allows for autonomous code building. The updates aim to improve the platform's ability to produce polished applications, but still require careful crafting of prompts.

PJM Interconnection, the largest US power grid operator, expects a 20% increase in electricity bills due to high demand from data centers and AI applications like ChatGPT, outpacing supply from closed fossil-fuel plants and delayed renewable energy projects.

Research by Forescout found that large language models (LLMs) are not effective in performing vulnerability discovery and exploitation tasks, with high failure rates across all models tested. The study analyzed 50 AI models and found that none were able to complete all tasks, with most models being unstable and requiring substantial user guidance.

Tesla has applied to test and operate autonomous vehicles in Arizona, aiming to expand its robotaxi service to the Metro Phoenix area. The company has begun the certification process and is expected to receive a decision by the end of the month. This move follows the launch of its limited robotaxi service in South Austin last month.

AWS is launching an AI agent marketplace next week with Anthropic as a partner, allowing startups to offer their AI agents to AWS customers and enabling enterprise customers to browse and install agents from a single location.

LGND is developing a tool to make geospatial data more accessible and efficient, using vector embeddings to summarize spatial data and make it easier to find relationships between different points on Earth. The company has raised a $9 million seed round and aims to take a bite out of the $400 billion geospatial market.

The data industry is consolidating due to AI adoption. Recent acquisitions, such as Databricks buying Neon and Salesforce acquiring Informatica, aim to improve data quality and management. However, experts question whether this strategy will achieve the desired AI-driven outcomes, as the acquired companies may not be optimized for the rapidly evolving AI landscape.

Anna Berenberg discusses One Network, a cloud-agnostic service and policy-oriented network architecture. It centralizes policy, simplifies operations, and enhances developer velocity. One Network is built on five principles: common foundation, everything as a service, unify all paths and support all environments, service extension open ecosystem, and apply and enforce uniform policies.

Cloudflare's CEO Matthew Prince has announced a new policy to block AI companies from scraping websites without compensating content creators, targeting Google's AI model Gemini. This move challenges the traditional web economy and sets a new standard for AI companies to negotiate with creators for access to their content.

Researchers have developed an AI agent system called A1 that can autonomously discover and exploit vulnerabilities in smart contracts, potentially stealing cryptocurrency. The system uses various AI models to generate exploits and has a high success rate, demonstrating the need for improved security measures in the cryptocurrency industry.

China's attempts to use AI for sign language translation on TV have been unsuccessful, with a professor noting that the technology loses key information and struggles to understand nuances of sign language, highlighting the need for greater involvement of sign language linguists and Deaf people in development