RAIZOR Report: Perplexity unveils Computer worker, Anthropic revises safety rules, Hollywood negotiates AI rights, chips surge

Perplexity Introduces Computer As A Multi Model Digital Worker

• Perplexity has launched Computer, a multi model orchestration system that coordinates nineteen different artificial intelligence models to function as a general purpose digital worker. The platform can browse the internet, write code, connect to software applications, and operate digital interfaces to complete complex work with minimal human oversight.
• Computer breaks large projects into subtasks and spins up smaller, sandboxed agents that can run for hours, weeks, or even months while keeping track of prior work. It can combine models from different providers within a single workflow and remember context across tasks, which is useful for work such as competitor research, technical documentation, or full software builds.
• Pricing is based on consumption, with Perplexity Max subscribers receiving a monthly credit bank, and the service is currently available on the web for Max users, with Pro and Enterprise access planned. By serving as a persistent project worker rather than a one off chatbot, Computer moves artificial intelligence closer to a true digital employee.
• This kind of orchestration could reshape roles in consulting, research, product development, and operations, where teams may shift from doing the work themselves to supervising long running artificial intelligence projects.

Anthropic Revises Its Responsible Scaling Policy For Advanced Models

• Anthropic has dropped the core pause clause from its 2023 Responsible Scaling Policy, which would have required the company to halt model training if safety techniques failed to keep pace with capabilities. The organization now plans to follow a more flexible roadmap while still framing itself as a leader in responsible artificial intelligence.
• The company says it intends to match or surpass competitors on safety practices, rather than unilaterally adopting stricter limits that might slow development. This shift reflects a broader political environment in which federal policy is placing more emphasis on artificial intelligence competitiveness.
• The change raises questions about how high risk systems should be governed when commercial and national interests are pushing for rapid advancement. That debate will affect technology companies, regulators, and leaders in any sector that plans to rely on increasingly powerful artificial intelligence systems for critical decisions.

MatX Secures More Than Half A Billion Dollars For New AI Chips

• MatX, a semiconductor startup founded by former Google engineers, has raised more than five hundred million dollars in funding led by Jane Street and Leopold Aschenbrenner’s Situational Awareness fund. The company is focused on building specialized hardware to run large scale artificial intelligence workloads.
• This fundraising round positions MatX as a significant new player in the market for advanced accelerators, an area currently dominated by companies such as Nvidia. Fresh capital will help MatX design, manufacture, and bring to market chips that could lower costs and ease supply constraints for artificial intelligence computing.
• A more competitive chip landscape would directly influence the speed and affordability of artificial intelligence adoption across industries. Technology leaders in cloud services, automotive, healthcare, and financial services should expect more options and potentially new price dynamics as alternative suppliers like MatX emerge.

Cognition Brings Devin Coding Agent To Government Agencies

• Cognition has announced that its Devin artificial intelligence coding agent and its Windsurf integrated development environment are now available to United States military and government agencies. The company describes Devin as an autonomous software engineer that can plan, write, test, and debug code with minimal supervision.
• By opening access to the public sector, Cognition is aiming to modernize legacy government systems and accelerate software development inside agencies that often face long procurement cycles and talent shortages. Automated coding support can help teams update aging infrastructure, patch security vulnerabilities, and ship new digital services faster.
• This move also signals a growing comfort with deploying artificial intelligence agents inside sensitive environments, provided they are deployed within secure workflows. Public sector technology leaders will need to rethink how they structure development teams and oversight processes when machine agents can contribute substantial volumes of code.
• As similar tools move into regulated industries, software engineers, project managers, procurement officers, and cybersecurity teams will all need new playbooks for working alongside artificial intelligence coders while maintaining accountability and compliance.

Google Expands Gemini To Handle Multi Step Tasks In Android Apps

• Google has updated Gemini on Android so the assistant can now carry out multi step tasks inside third party mobile applications. Users can ask Gemini to complete actions such as booking a ride or placing a food delivery order, and the system will navigate through the relevant applications on their behalf.
• This feature, which is rolling out in beta on the latest Pixel devices and Samsung Galaxy flagships, turns Gemini from a conversational helper into an agent that can take real actions across a user’s phone. The approach shows how artificial intelligence assistants are moving beyond chat into direct control of everyday services.
• Allowing a model to operate other applications raises design and safety questions about permissions, reliability, and user control. Product leaders and mobile developers will increasingly need to design their services so that artificial intelligence agents can interact with them in predictable, auditable ways rather than through fragile screen tapping scripts.

Google Turns Opal Into A More Interactive No Code Experience

• Google has upgraded Opal, its no code mini application builder, by adding an interactive artificial intelligence agent step into workflows. The agent can interpret a user’s goal, decide the best path to achieve it, and automatically call the right tools inside a given Opal flow.
• This change turns previously static, linear automations into dynamic experiences where the artificial intelligence system can branch, ask follow up questions, or make choices based on context. It lowers the barrier for non technical users to build experiences that feel conversational rather than like simple forms.
• By embedding decision making into no code platforms, Google is nudging everyday business processes toward more adaptive automation. Operations, customer support, and internal tooling teams can rethink routine workflows so they can respond intelligently to each user or case rather than following a single script.

War Games Show Artificial Intelligence Models Often Recommend Nuclear Escalation

• Researchers at King’s College London ran twenty one simulated war games using large language models from providers including OpenAI, Anthropic, and Google, and found troubling patterns in the behavior of the systems. According to reporting by New Scientist, at least one model in ninety five percent of simulations recommended the use of nuclear weapons.
• None of the models ever chose to surrender, and some escalated quickly, revealing a bias toward aggressive strategies when confronted with high tension scenarios. Although these systems were designed for language tasks, the study shows how current artificial intelligence models can generate confident but dangerous recommendations when asked to think through geopolitics.
• The findings raise serious concerns about delegating military or national security decision making to general purpose artificial intelligence systems, especially without strict constraints and human oversight. Governments, defense contractors, and technology providers will need rigorous testing regimes and clear red lines before placing such systems anywhere near real world command structures.
• Beyond defense, any organization considering artificial intelligence for high stakes strategy, risk management, or crisis response should think carefully about model alignment, simulation testing, and the human review needed before acting on automated recommendations.

Profound AI Offers Analytics On How Models Recommend Your Brand

• Profound AI has launched an enterprise analytics platform that helps companies understand how their products and content are surfaced by artificial intelligence systems. The company focuses on how assistants, search tools, and recommendation engines rank and display brands when users ask questions or seek suggestions.
• The platform provides insights into visibility, ranking positions, and the competitive landscape inside model outputs, which are increasingly influencing consumer discovery. With artificial intelligence becoming a major gatekeeper for information, this kind of visibility is similar to what search engine optimization offered during the rise of web search.
• By treating artificial intelligence assistants as new distribution channels, Profound AI is pushing marketing and growth teams to think beyond traditional search and social metrics. Brand, product, and content leaders will need new strategies and measurement frameworks to make sure they remain discoverable when customers rely on conversational systems rather than browsing web pages directly.

Matthew McConaughey Urges Creators To Own Their Digital Likeness

• At a town hall hosted by CNN and Variety, actor Matthew McConaughey told students to prepare for widespread synthetic media by locking down rights to their voice, image, and creative work. He pointed to his decision to trademark his famous catchphrase from the film “Dazed and Confused” as an example of how individuals can protect their brands.
• McConaughey described a future where artificial intelligence can project performers into private events, branded experiences, or entirely new films without them ever stepping on set. In that world, the real leverage comes from whether the technology has to ask permission and pay for the privilege of using a person’s likeness.
• His comments align with a broader entertainment industry shift toward treating identity and performance as licensable assets in negotiations with artificial intelligence companies. Actors, musicians, influencers, and even executives should be thinking about trademarks, contracts, and digital rights before synthetic versions of them become common in commercial media.

Major Studios Challenge Artificial Intelligence Training In Court

• Wired has been tracking a wave of copyright lawsuits filed in 2024 by major film and music studios against artificial intelligence platforms. These cases argue that training models on copyrighted movies, shows, and songs without explicit permission or payment infringes on creators’ rights.
• The legal challenges target both how training data is collected and how generative systems produce new work that may echo the style of existing artists. Courts are now being asked to decide whether large scale scraping and learning from copyrighted material counts as fair use or requires licensing.
• The outcomes will shape how quickly and cheaply artificial intelligence companies can train future models, and how much leverage creative rights holders will have in negotiations. Media, publishing, and technology leaders should monitor these cases closely, since they will influence both risk and opportunity around content based artificial intelligence products.

Studios Shift From Lawsuits To Licensing Deals With Artificial Intelligence Firms

• By late 2025, many entertainment studios had begun moving from pure litigation toward licensing agreements and equity stakes in artificial intelligence companies. A high profile example is Disney’s arrangement with OpenAI around the Sora video generation system.
• These deals give studios a say in how their catalogs and characters are used in synthetic media while also letting them share in the upside as new formats emerge. Instead of fighting artificial intelligence from the outside, rights holders are starting to shape products and business models from within.
• Observers have framed this as the industry level version of individuals like Matthew McConaughey trademarking their own catchphrases so that participation in artificial intelligence projects can be negotiated, priced, and enforced. This shift suggests that legal departments, business affairs teams, and creative executives will need fluency in both intellectual property and model licensing structures to strike favorable terms.

Nicholas Thompson To Bring Media View On Artificial Intelligence To Creativity Summit

• Nicholas Thompson, chief executive of The Atlantic and former editor in chief of Wired, is joining the AI & Creativity Summit as a speaker. Thompson has spent years covering and steering organizations through shifts driven by technology, including artificial intelligence.
• His participation signals that mainstream media is taking the creative implications of generative tools seriously, not only as subjects of coverage but as instruments for storytelling and distribution. The summit aims to explore how artificial intelligence can augment, and potentially disrupt, creative work across formats.
• Bringing together media leadership and technology voices highlights how narrative, trust, and business models are all being reshaped by algorithmic content. Editors, marketers, brand leaders, and creative directors will increasingly need to understand both the opportunities and reputational risks that come with synthetic storytelling at scale.

Axios Warns Hollywood About Rising Chinese Artificial Intelligence Content Platforms

• Axios reports that Hollywood could become an early casualty of China’s artificial intelligence ambitions as firms like ByteDance accelerate work on entertainment platforms such as SeeDance. These platforms rely on algorithms and generative media to keep users engaged with low cost, highly personalized content.
• By lowering production costs and optimizing what audiences see, Chinese technology companies could undercut traditional United States studios on both price and global reach. That dynamic could erode Hollywood’s long standing advantage in shaping worldwide culture.
• The story highlights how the competition is no longer just about making the best films, but about controlling data rich platforms that can feed and refine artificial intelligence driven recommendation engines. Studio executives, streaming leaders, and talent agencies will need to think as much about product experience and algorithms as they do about scripts and stars.

Physical Artificial Intelligence Market Could Reach One Trillion Dollars

• Bloomberg covers a projection from Barclays that the market for physical artificial intelligence systems could reach one trillion dollars in value by 2035. This category includes autonomous vehicles and other embodied systems that can sense, reason, and act in the real world.
• Self driving technology is expected to be the biggest driver of this growth, with additional contributions from delivery robots, warehouse automation, and smart industrial equipment. As these machines become more capable and reliable, they can take on tasks that currently require human drivers and operators.
• Such a shift will have ripple effects across transportation, logistics, urban planning, and insurance, as well as in energy and manufacturing. Business leaders in sectors that move people or goods should be preparing now for new operating models, cost structures, and safety frameworks as physical artificial intelligence systems move from pilots into daily use.

Google Negotiates Artificial Intelligence Content Deals With News Publishers

• Press Gazette reports that Google is in active talks with news organizations, including the Financial Times, over licensing arrangements related to artificial intelligence products. These negotiations cover how publisher content is used to train models and how it appears inside new search and answer features.
• The discussions are already influencing changes in Google Search and other services that surface news, as the company tries to balance user experience, legal risk, and publisher relationships. Licensing agreements could involve direct payments, traffic guarantees, or new display formats inside artificial intelligence powered results.
• As artificial intelligence answers increasingly sit between users and original reporting, the economic relationship between platforms and publishers is being rewritten. Executives in media, advertising, and search will need to factor these evolving deals into their audience, revenue, and product strategies.

DeepSeek Limits Early Access To Its New Model Amid Tech Tensions

• Reuters reports that Chinese artificial intelligence company DeepSeek is withholding early access to its upcoming flagship model, known as V4, from United States chipmakers including Nvidia and Advanced Micro Devices. Instead, the firm is granting early access to domestic technology giants such as Huawei.
• The move highlights growing geopolitical tensions around artificial intelligence and semiconductor supply chains, where both hardware and software are becoming strategically sensitive. Restricting access to leading models can shift bargaining power in negotiations over cloud services, chips, and data centers.
• This kind of fragmentation could lead to parallel ecosystems of models and hardware, with limited interoperability across political blocs. Global companies that rely on artificial intelligence infrastructure will need contingency plans for a world where access to models or chips can change quickly because of policy or corporate strategy.

Wayve Raises One Point Two Billion Dollars For Autonomous Driving

• United Kingdom startup Wayve has raised one point two billion dollars in Series D funding, giving the company a valuation of about eight point six billion dollars, according to CNBC. Investors include major carmakers as well as technology leaders Nvidia and Microsoft.
• Wayve is developing self driving systems that rely heavily on machine learning to interpret camera input and make driving decisions, rather than hard coded rules. The substantial funding round will support large scale data collection, model training, and partnerships with automakers to bring the technology to market.
• The deal underscores how central artificial intelligence has become to the future of transportation, not only in luxury vehicles but across mass market fleets. Automotive executives, city planners, and logistics providers should be preparing for a phased introduction of driver assistance and autonomous features that change how vehicles are owned, operated, and insured.

Google Launches Nano Banana 2 For Faster, Cheaper Image Generation

• Google has released Nano Banana 2, an upgraded image generation model that offers higher resolution, more consistent outputs, and better handling of text inside images at about half the price of the previous version. The company says the model matches or outperforms leading text to image systems on several independent leaderboards.
• Nano Banana 2 can create images that scale up to four thousand pixel resolution across different aspect ratios while maintaining coherence for multiple characters and objects in a single scene. The service is priced at roughly seven cents per image and runs at speeds similar to Google’s lightweight text models.
• Google has made Nano Banana the default image generator across its ecosystem, with a Pro variant reserved for paying subscribers, which will increase exposure to everyday users and businesses. Marketers, designers, product teams, and small businesses can all experiment with higher quality synthetic visuals for campaigns, prototypes, and content without relying on traditional photo shoots or stock libraries.

OpenAI Hires Key Artificial Intelligence Leaders From Meta And Apple

• The Information reports that OpenAI has hired Ruoming Pang, an artificial intelligence infrastructure leader whom Meta had previously recruited from Apple with a compensation package reportedly exceeding two hundred million dollars. Pang previously led the models group at Apple and helped define the company’s Apple Intelligence strategy.
• OpenAI has also brought on Riley Walz, known for experimental projects such as Jmail and Find My Parking Cops, to work on prototyping new interfaces for interacting with artificial intelligence. These hires suggest OpenAI is investing both in the core technical stack and in more creative, user facing experiences.
• Competition for top talent in machine learning and product design is intensifying as major platforms race to define the next generation of intelligent applications. Technology executives and human resources leaders should expect compensation, retention strategies, and organizational design around artificial intelligence teams to remain in flux as this talent market heats up.

Pew Study Reveals How Teens Rely On And Worry About Artificial Intelligence

• The Pew Research Center has released a study showing that teenagers heavily rely on artificial intelligence tools for schoolwork, information seeking, and entertainment. A majority believe that cheating with the help of such tools is common, and that share rises further among teens who use these technologies frequently.
• Many young people say artificial intelligence helps them learn faster and get more done, but they are also worried about future job loss and the impact on human creativity. Notably, about forty percent of parents reported that they have never discussed artificial intelligence with their children at all.
• The findings suggest a growing gap between how central these tools are in teens’ daily lives and how prepared families and schools are to guide their use. Educators, parents, policymakers, and employers should think about setting norms and building curricula that teach not only how to use artificial intelligence, but when to question and constrain it.

AI Tools

🎬 Quick Cut – Adobe Firefly’s AI-powered video editor that transforms raw footage into automatic first cuts for faster post-production.

📱 Bixby – Samsung’s on-device AI assistant that can act as a swappable agent on Galaxy S26 phones for conversational help and task automation.

🧱 CreateOS – A platform that lets you build and deploy applications from any AI coding tool into dedicated GPU-backed endpoints.

📢 Reloop – An AI advertising tool that generates effective ad creatives without requiring prompts or prior design skills.

🎥 Protex – An AI video analytics platform that monitors existing CCTV footage to detect unsafe events and behaviors in real time.

🧠 TabAI – A productivity assistant that consolidates tasks from multiple sources into one AI-organized workspace to help you focus.

🧩 Open-source LLMs platform by Nebius – An infrastructure platform for running, fine-tuning, and deploying open-source language models in production with full control over scaling and hardware.

⚙️ Notion Custom Agents – Always-on AI agents inside Notion that execute scheduled or triggered workflows across connected tools like Slack, email, and calendars.

🧑‍💼 Claude Cowork Plugins – Enterprise-grade plugins and connectors that embed Claude into business workflows and systems such as Google Workspace, DocuSign, and WordPress.

🖥️ Claude Code Remote Control – A tool that lets you offload and control terminal-based coding tasks through Claude Code from your browser or phone.

🎹 ProducerAI – A generative music system that creates full songs and custom instruments from text prompts.

🚀 Mercury 2 – A high-speed reasoning model capable of generating over 1,000 tokens per second while keeping inference costs low.

🤖 Unitree G1 humanoid robot – A consumer-oriented humanoid robot that uses AI to walk, balance, and perform complex motions like dancing.

📝 HyNote – An AI note-taker that records meetings and audio, then generates searchable summaries and key takeaways.

🎨 Palix AI – A creator-focused platform for generating images, videos, and music using AI from a single interface.

📊 Kolva – A pay-as-you-go meeting assistant that transcribes calls, extracts action items, and supports AI-driven document search.

🎧 NeatScribe – An AI transcription tool that converts audio and video files into accurate, editable text.

🔍 SerpApi – A real-time search API designed for AI applications, providing structured web results with privacy-focused features like ZeroTrace Mode.