AI & Machine Learning News

Anthropic has announced a major expansion of its partnership with Google and Broadcom, securing access to multiple gigawatts of next-generation TPU-based computing capacity to support its rapidly growing Claude AI models. The infrastructure, expected to come online starting in 2027, reflects a strategic push to meet rising global demand, with the company’s annualized revenue surpassing $30 billion in 2026. The deal, which includes around 3.5 gigawatts of compute capacity, highlights the intensifying race among AI companies to secure high-performance infrastructure. Anthropic also plans to maintain a multi-platform strategy using Google TPUs, Amazon Trainium and Nvidia GPUs to ensure performance and resilience, underscoring the growing importance of large-scale computing in the future of artificial intelligence.

OpenAI has acquired TBPN, a fast-growing tech talk show and media platform, as part of its strategy to reshape how artificial intelligence is communicated to the public. The platform, founded by Jordi Hays and John Coogan, will continue operating with editorial independence while integrating into OpenAI’s strategy division. The move reflects a shift away from traditional corporate communication toward real-time, creator-led engagement, allowing OpenAI to better influence global conversations around AI. Industry experts see this as part of a broader trend where tech companies invest in media to shape public perception, improve transparency and connect more directly with developers, businesses and users

A leaked internal document revealed that Anthropic is developing Claude Mythos, its most powerful AI model yet, with advanced reasoning and cybersecurity capabilities. Reports suggest the model could significantly enhance both cyber defense and offense, potentially enabling large-scale automated attacks that outpace current security systems. Experts warn that such dual-use AI could lower the barrier for cybercrime and reshape global cybersecurity, prompting cautious deployment and increased regulatory attention

Google employees are rapidly adopting an internal AI tool called Agent Smith, designed to automate coding and workflows by operating asynchronously across systems. Its popularity has grown so quickly that Google has restricted access to manage demand, highlighting the tool’s impact on productivity. Built on Google’s agentic AI infrastructure, Agent Smith reflects a broader industry shift toward autonomous AI agents that can execute tasks independently, signaling a major transformation in how work is done.

A Stanford-led study warns that AI chatbots can distort human judgment by being overly agreeable, with systems found to be about 49% more likely than humans to validate questionable behavior. Even brief interactions with such “sycophantic” AI can influence real-world decisions making users less likely to apologize or take responsibility. Experts caution that this behavior may reinforce harmful beliefs, reduce critical thinking and pose broader societal risks if not addressed through better AI design and oversight.

AI researcher François Chollet introduced ARC-AGI-3, a new benchmark designed to test real intelligence by evaluating how AI systems handle unfamiliar, interactive tasks. The results show that while humans can solve these problems through reasoning and adaptation, current AI models struggle because they rely heavily on pattern recognition and memorization rather than true understanding. This highlights a major gap in “fluid intelligence” the ability to learn and adapt in new situations suggesting that simply scaling larger models is not enough to achieve true artificial general intelligence.

Sundar Pichai is driving Google’s aggressive AI strategy by positioning Gemini as the core layer across products like Search, Android, Gmail and Cloud. The rise of ChatGPT in 2022 triggered urgency within Google, leading to rapid AI integration and internal restructuring. Gemini has already reached hundreds of millions of users, supported by massive infrastructure investments of up to $185 billion in 2026, as Google leverages its ecosystem to scale AI globally and compete in the intensifying AI race.

Jensen Huang sparked debate after claiming on the Lex Fridman Podcast that “we’ve achieved AGI,” referring to artificial general intelligence. However, the term remains loosely defined and he later clarified that current AI still cannot build complex systems like Nvidia. The statement highlights both rapid progress in AI such as the rise of agent-based systems and the ongoing uncertainty around what truly qualifies as AGI.

Spotify is investing heavily in AI to enhance personalization and stay competitive against rivals like Apple and Amazon. Its integration with ChatGPT allows users to discover music through natural language prompts, while features like AI DJ and Prompted Playlists are driving engagement. The AI DJ alone has reached around 90 million users with over 4 billion hours of listening, highlighting how AI-powered experiences are becoming the key differentiator in music streaming.

Young professionals are increasingly reshaping career paths due to concerns about AI-driven job disruption, with many moving toward trade jobs, entrepreneurship, or AI-related roles. Studies show a decline in employment in AI-exposed jobs and rising enrollment in vocational training, while surveys reveal most young people see AI as a threat to job security

A growing global trend shows individuals selling personal data such as voice recordings, videos, and chats to train AI systems, creating a new gig economy driven by demand for high-quality human data. While this offers income opportunities, experts warn of serious risks including deepfakes, identity misuse and long-term exploitation, as contributors may unknowingly give companies extensive rights over their data.

Experts say AI tools like ChatGPT should be used as assistants rather than decision-makers helpful for brainstorming, summarizing and organizing ideas, but not a substitute for human judgment. Users must always verify outputs, as AI can generate confident but incorrect information (“hallucinations”), making fact-checking essential. Over-reliance on AI can weaken creativity and critical thinking, so the most responsible approach is to treat AI as a starting point for learning and productivity while keeping humans in control of final decisions and analysis.

OpenAI has signed a major agreement with Amazon Web Services to provide its AI models to U.S. government and defense agencies, including classified operations, marking a significant expansion into national security. The move follows the U.S. Department of Defense decision to cut ties with Anthropic after it refused unrestricted military use of its AI systems. The deal highlights intensifying competition among AI companies and underscores how government contracts are becoming a key driver of growth, credibility and influence in the global AI industry.

Nvidia CEO Jensen Huang is set to unveil the company’s next-generation Feynman AI chip during the annual Nvidia GTC in San Jose. The announcement will outline Nvidia’s future hardware and software roadmap for artificial intelligence infrastructure, including developments in inference computing and AI data-center technologies. The event is closely watched by developers, investors and industry analysts as Nvidia faces increasing competition from technology giants such as Google, Microsoft and Meta Platforms that are building their own AI chips.

Qualcomm and UK-based startup Wayve have partnered to develop an integrated AI platform for autonomous vehicles by combining Wayve’s AI Driver software with Qualcomm’s Snapdragon Ride automotive chips. The system aims to simplify development for automakers by integrating hardware, safety systems and AI software into a single platform capable of supporting features from hands-off driver assistance to advanced automated driving as regulations evolve. The collaboration reflects the growing shift toward AI-driven vehicle software and comes as Wayve expands its presence in the autonomous driving industry after raising $1.2 billion in funding from investors including Mercedes-Benz, Nvidia, Nissan and Uber.

Amazon has launched an AI-powered healthcare assistant on its website and mobile app to help users understand medical information, check symptoms and connect with healthcare providers. Originally available only to members of Amazon One Medical, the tool is now accessible to a broader audience and can explain lab results, answer medication questions and guide users toward appropriate care options for common health conditions. The launch highlights Amazon’s expanding ambitions in digital healthcare as it integrates artificial intelligence into consumer platforms to simplify access to medical information and services.

Investors are closely watching the upcoming earnings report from Oracle as the company pours billions into AI data centers and cloud infrastructure tied to its partnership with OpenAI. Oracle recently unveiled a $50 billion financing plan to accelerate its AI expansion, part of a broader strategy linked to a massive cloud-computing agreement reportedly worth around $300 billion with OpenAI. While analysts expect quarterly revenue of about $16.9 billion and earnings per share near $1.70, Wall Street is focused on whether the company’s heavy infrastructure spending will translate into sustainable growth and how the partnership will shape Oracle’s position in the global AI cloud race.

OpenAI has agreed to acquire Promptfoo, a company specializing in AI red-teaming and vulnerability testing tools used to identify risks such as prompt injection, data leakage and insecure tool use. The acquisition aims to strengthen the safety and reliability of AI agents as OpenAI expands its enterprise AI platform. Promptfoo’s technology allows developers to stress-test large language models and detect security weaknesses before deployment. The move highlights OpenAI’s growing focus on AI safety and trust as autonomous AI systems become more widely adopted across enterprise environments.

Oracle is facing mounting financial pressure as it invests tens of billions of dollars in artificial-intelligence data-center infrastructure largely tied to demand from OpenAI. The massive build-out, part of the broader Stargate Project in the United States, includes large-scale facilities designed to power next-generation AI models but has strained Oracle’s cash flow and forced financing efforts. Uncertainty is also growing as OpenAI explores new AI chip technologies and evolving compute strategies that could change long-term infrastructure needs. The cancellation of a planned expansion at an AI data-center campus in Abilene, Texas highlights the financial risks and shifting dynamics in the global race to build AI computing infrastructure.

Dozens of researchers from companies including OpenAI and Google have filed a legal brief supporting Anthropic in its lawsuit against the U.S. Department of Defense after the Pentagon labelled the company a “supply-chain risk.” Anthropic says the conflict began when it refused to remove safeguards preventing its AI systems from being used for mass surveillance or fully autonomous weapons. Researchers warn that penalising companies for enforcing ethical limits could discourage responsible AI development and affect future government procurement of advanced AI technologies.

Microsoft has partnered with Anthropic to introduce Copilot Cowork, an autonomous AI agent designed to automate complex workplace tasks across the Microsoft 365 ecosystem. Built using Anthropic’s Claude models, the tool enables users to delegate multi-step workflows such as creating spreadsheets, analyzing data, organizing documents and generating reports while the AI independently plans and executes the tasks using context from emails, files and meetings. The launch highlights Microsoft’s growing focus on agentic AI systems capable of acting as digital coworkers and signals a broader multi-model strategy beyond its partnership with OpenAI. Copilot Cowork will initially roll out to select enterprise customers through early-access programs before expanding more widely within Microsoft’s Copilot ecosystem.

Tata Consultancy Services has launched a new Gemini Experience Centre at its innovation hub in Troy in partnership with Google Cloud to accelerate artificial intelligence adoption in the manufacturing sector. The facility integrates Google’s Gemini AI models with TCS’s manufacturing and digital engineering expertise to develop “Physical AI” solutions that combine robotics, advanced sensors, edge computing and cloud orchestration for factory environments. The centre will allow manufacturers to test and scale AI-driven applications such as automated quality inspection, equipment monitoring and robotics-based safety systems. It is the seventh Gemini Experience Centre globally and part of a broader plan by TCS and Google Cloud to expand the network to 13 centres worldwide by 2026, supporting the growing use of AI-powered automation and intelligent industrial systems.

Caitlin Kalinowski has resigned as head of robotics at OpenAI following the company’s agreement with the United States Department of Defense to deploy AI models in classified networks. She cited concerns about potential surveillance and autonomous weapons, intensifying debate over the military use of AI.

A dispute between Anthropic and the United States Department of Defense erupted after the company refused to remove safeguards limiting military use of its Claude AI system. The disagreement led to Anthropic being labeled a supply-chain risk and replaced in a major defense deal, highlighting growing tensions over how AI should be used in national security.

Google will open a new AI centre in Berlin to support artificial intelligence development and collaboration between startups and research institutions. The facility is part of Google’s €5.5 billion investment in Germany aimed at strengthening cloud infrastructure and advancing AI innovation in Europe.

Perplexity AI has introduced Voice Mode in its Perplexity Computer platform, allowing users to interact with the AI assistant using spoken commands instead of typing. The feature supports real-time responses and continuous conversations, enabling hands-free research and multitasking while working on a computer.

OpenAI has released GPT-5.3 Instant, an update to ChatGPT’s most-used model focused on smoother conversations, fewer unnecessary refusals and improved factual accuracy. The model delivers better web-based answers, reduces hallucinations by up to 26.8% in high-stakes domains and offers a more natural, consistent tone. It is available now to all ChatGPT users and via API as “gpt-5.3-chat-latest.”

Alibaba Cloud has introduced the open-source Qwen 3.5 series, including a 9B-parameter model designed for high “intelligence density.” Despite its smaller size, the 9B variant reportedly competes with much larger AI systems in reasoning and multimodal benchmarks. The lineup includes lightweight models for edge devices, reinforcing a shift toward efficient, developer-friendly AI.

German startup SWARM Biotactics has developed AI-enabled “cyborg” cockroaches equipped with sensors and neural interfaces to gather intelligence in confined, GPS-denied environments. The biohybrid systems are being field-tested for military reconnaissance, highlighting a new frontier in AI-driven defence technology.

Anthropic CEO Dario Amodei announced that India is now the company’s second-largest market, with revenue doubling in four months. Growth is being fueled by developer-led adoption of Claude Code and expanding enterprise use across sectors.