The day's AI news, briefed
The Daily Prompt
Our AI editor reads each day's AI and robotics stories, pulls out why they matter, and links every original source — so you can keep up in minutes.
Join over 250,000 subscribers getting the latest AI news
Latest AI & Robotics News Today
Never use ChatGPT for these five tasks
Using ChatGPT for sensitive tasks can lead to severe legal and personal risks, emphasizing the need for users to understand its limitations regarding privacy and privilege.
- ChatGPT is not a substitute for professional advice; it lacks legal privilege, meaning conversations can be disclosed in legal contexts.
- Over 2,000 court cases have been hindered by AI-generated false citations and information, impacting both self-represented litigants and lawyers.
- Uploading confidential documents to ChatGPT poses data security risks, as illustrated by a former DHS official's breach of sensitive material.
- Users have attempted to rely on ChatGPT for medical advice, risking health outcomes due to a lack of proper safeguards for sensitive conversations.
- ChatGPT conversations can be indexed and discovered publicly, raising concerns over personal privacy and identity theft.
ChatGPT is an AI chatbot developed by OpenAI that generates human-like text responses. Due to previous incidents involving misuse and lack of confidentiality, OpenAI updated its policies to prohibit using the service for legal and medical advice without professional supervision. Users should be aware of the risks involved with sharing sensitive information with AI services.
Read the original ↗
One to More, More to One: Category-Aware Iterative Expert Training for Software Engineering Agents
The development of category-aware training for software engineering agents improves their efficiency, showcasing advancements in reinforcement learning applications within the software development domain.
- The proposed framework uses category-aware expert training to manage heterogeneous task categories in software engineering agents.
- It achieves a performance of 58.04% on Pro-618 and 59.00% on SWE-bench Multilingual without external solution-trajectory teachers.
- The Refresh–Repair–Expand (RRE) policy iteratively develops category experts, while the label-routed multi-teacher on-policy distillation (MOPD) consolidates them into a deployable model.
This research focuses on enhancing software engineering agents through advanced reinforcement learning methods. Previous models faced challenges with task-specific progress, which this new model aims to address through improved training frameworks. Future advancements may include further refinements and applications of this technology in real-world software engineering tasks.
Read the original ↗
Why Agentic Payments Will Be the Ultimate Test of AI Autonomy
The development of agentic payment systems will determine the extent to which AI can autonomously manage financial transactions without human intervention, impacting economic risks and consumer trust.
- Agentic payments shift AI's role from executing transactions to determining what it should autonomously handle, raising new governance challenges.
- Companies like Visa, Mastercard, and Stripe are advancing the infrastructure for agentic commerce, but governance frameworks lag behind.
- Effective governance for agentic payments should be based on potential consequences of errors, reversibility of actions, and customer context.
Agentic payments represent a significant evolution in AI autonomy, moving from suggestion-based systems to those that can autonomously manage financial transactions. Understanding how AI can make decisions on payments without constant human oversight is becoming crucial as this technology matures. Stakeholders must ensure a balance between autonomy and necessary governance to mitigate risks.
Read the original ↗
AI Promised to Organize My Clutter. Here’s the Messy Result
The effectiveness of AI-driven organizational tools like SmartLabels can significantly impact home efficiency, particularly in managing cluttered spaces, which many households face.
- SmartLabels consists of QR-code stickers and a companion app that helps keep a digital inventory of stored items in bins or drawers.
- Accuracy of the SmartLabels system relies on user diligence in updating the app with changes to item locations or additions.
- The app generates an itemized inventory from photos but may misidentify some products if not properly cataloged by users.
- SmartLabels offers 48 unique QR-code stickers and a free app, with a premium subscription option for additional features.
- The system can be useful for infrequently accessed items like keepsakes, but requires consistent updates for everyday storage solutions.
SmartLabels is a digital organization system aimed at helping users keep track of their belongings using QR codes and an app. Previously, managing household clutter often relied on physical labels and memory. Users should monitor how effective such AI solutions are in clutter prevention as they gain popularity.
Read the original ↗
AI Agents Are Becoming a New Malware Distribution Channel
The emergence of AI agents as malware distribution channels poses significant risks to cybersecurity, as they can mislead users into executing malicious software without their awareness.
- FakeGit, a malware campaign, involved 7,600 fake GitHub repositories and over 14 million downloads, distributing infostealers through impersonated AI skills.
- Gemini and ChatGPT independently recommended the same fake repository, illustrating how agents can unintentionally promote malicious content.
- Attackers exploit agent vulnerabilities by embedding deceptive instructions in tool descriptions, allowing agents to carry out unauthorized actions without user knowledge.
AI agents are computer programs that assist users by processing instructions, but recent findings indicate they may unintentionally facilitate malware distribution. The FakeGit campaign showcases this evolving threat by using AI suggestions to spread malicious software. Future developments should focus on enhancing the security of AI systems to mitigate these risks.
Read the original ↗
Ema raises $77M as AI starts eating into enterprise software and services
The rise of AI in enterprise software signals a pivotal shift in how businesses operate, potentially reducing reliance on traditional software solutions and reshaping service models.
- Ema has raised $77 million in a Series B funding round, bringing its total funding to $140 million.
- The startup uses AI agents to automate processes in HR, IT, and finance, aiming to replace traditional enterprise software.
- With over 50 enterprise customers, including Google and Microsoft, Ema's revenue bookings have surpassed $150 million, growing 50-fold in the past two years.
- Ema is expanding its presence in new markets, targeting Asia-Pacific, South America, and parts of the Middle East over the next year.
Ema, founded in 2023, focuses on automating corporate processes with AI technology, attracting significant investment in a competitive market for enterprise software. As AI increasingly impacts business operations, Ema is positioned to innovate traditional IT service models and expand its global footprint.
Read the original ↗
AI Won’t Replace Accountants But It Will Change the Job
The introduction of AI in accounting addresses staffing shortages while enhancing productivity, potentially reshaping the industry's workforce structure without leading to significant job losses.
- A global survey showed 73% of accounting firms are turning away clients due to a staffing shortage.
- AI adopters reported a 55% increase in weekly client work, allowing accountants to take on more complex tasks.
- The Bureau of Labor Statistics projects around 115,300 annual openings for accountants through 2035, largely due to retirements.
- AI helps accountants move from repetitive tasks to critical thinking and client communication roles, improving overall efficiency.
The accounting industry is experiencing a staffing shortage, exacerbated by retiring professionals and a preference for STEM careers. AI is being integrated to boost productivity and support an aging workforce. Moving forward, the accounting sector must adapt to new roles and skills as technology evolves.
Read the original ↗
What Is Multimodal AI? How Models Combine Text, Images, Audio, and Video
Understanding multimodal AI is crucial for accurately assessing its capabilities and limitations, as well as for distinguishing it from simpler single-modality systems.
- Multimodal AI can handle various data types, including text, images, audio, and video, by relating, receiving, or generating information across these media.
- The article outlines a five-stage operating map for multimodal AI, detailing key processes from encoding input features to generating outputs in the requested medium.
- A major distinction of multimodal AI is its ability to connect and reason over multiple contexts, unlike separate single-modality tools.
Multimodal AI refers to systems that integrate information from diverse media for analysis and output. Previous developments primarily focused on single modalities, but the increasing demand for more complex interactions prompts advancements in multimodal capabilities. Future discussions should emphasize applications and improvements in integrating these systems.
Read the original ↗
Meta Says OpenClaw Inspired Its New AI App Muse
Meta's admission of drawing inspiration from OpenClaw highlights its strategy of leveraging successful, innovative features from popular AI applications to develop its own tools for a broader audience.
- Meta's new AI app, Muse, is heavily inspired by OpenClaw, according to Nat Friedman, head of product at Meta’s Superintelligence Labs.
- Friedman noted that while Muse draws from OpenClaw, it was built entirely from scratch by the Meta team.
- Muse has quickly gained popularity, reaching the number one spot on the U.S. App Store and outperforming ChatGPT's early mobile launch.
- OpenClaw, which became popular as a personal AI assistant, is an open-source tool that Meta studied for effective design and functionality.
Meta is a technology company known for its social media platforms and AI applications. The company has a history of incorporating successful features from other platforms into its own products. The development of Muse comes as AI applications continue to see rapid growth and adoption.
Read the original ↗
Anthropic Releases Claude Opus 5.5: Fable 5.1-Level Performance at 40% Lower...
Claude Opus 5.5's significant cost reduction and improved performance could enhance efficiency for developers and businesses relying on AI for coding and knowledge work, potentially reshaping the competitive landscape.
- Anthropic has released Claude Opus 5.5, which performs at the level of Claude Fable 5.1 but at 40% lower running costs than Opus 5.
- The model utilizes less compute and features a 60% reduction in cache reading costs, contributing to its decreased operating expenses.
- Opus 5.5 is deployable as a managed API model, available on platforms like AWS, Google Cloud, and Microsoft Azure.
- Tests show that Opus 5.5 completes coding tasks significantly faster than its predecessor, Opus 5, with multiple benchmarks indicating superior performance.
- A notable case involved one tester completing a 680,000-line code migration in under a day, a task Opus 5 had taken over 20 hours to finish.
Anthropic is an AI research organization focused on developing advanced models for various applications. Claude Opus 5.5 is the latest release in their line, following Opus 5, and featuring improvements in performance and cost efficiency. Observers will monitor its adoption and impact on competitive AI services.
Read the original ↗
SpeakON Ships a MagSafe AI Voice Button With Its Own Microphone
SpeakON's device offers a new hardware solution for voice input, addressing longstanding limitations in audio output for mobile dictation, which can enhance productivity for busy professionals.
- SpeakON has launched a 25 g MagSafe AI voice button that captures speech and inputs text directly into any active app on iPhones.
- The device features its own microphone and battery, allowing it to work offline and avoid common permission issues with background audio.
- SpeakON supports various output modifications including Smart Polish, Smart List, and multi-language translation, aiming for more polished text than traditional dictation.
- Priced at $129, the SpeakON device is designed for iPhone models with MagSafe technology and offers a one-time purchase without subscriptions.
SpeakON's AI Communicator is designed for professionals who need quick and efficient voice-to-text capabilities without the hassle of dictation cleanup. Traditionally, phone dictation systems have struggled with output quality and app switching, creating inefficiencies. Upcoming functionalities, like SpeakON Agent, will extend its capabilities further in October 2026.
Read the original ↗
OpenAI Releases GPT-6 Sol and Luna: 50% Cheaper API Pricing and...
The launch of GPT-6 Sol and Luna represents a significant reduction in costs for developers, potentially increasing accessibility and adoption of advanced AI capabilities across various applications.
- OpenAI has launched two new models, GPT-6 Sol and GPT-6 Luna, now available in its API as of September 2026.
- The models are priced at 50% lower than GPT-5.6, with Sol costing $2 per million tokens and Luna priced at $0.10 per million tokens.
- Luna provides a faster, cost-effective alternative for high-volume tasks, while Sol is designed for complex coding and professional applications.
- Both models utilize improved caching and inference methods, which allow OpenAI to offer these lower prices.
- Sol’s output price has dropped by about 58%, from $1.20 to $0.50 per million tokens.
The GPT-6 family is OpenAI's latest offering in its series of AI language models, following the earlier release of GPT-6 Astra. These developments build on advancements in AI model training techniques aimed at enhancing performance and reducing costs. The pricing and performance metrics suggest wider usage in professional and everyday tasks can be expected as these models roll out.
Read the original ↗
Kyutai Releases Voice of Reason: A Speech-Native Model that Solves Spoken...
The development of Kyutai's Voice of Reason offers significant advancements in speech-native models for solving spoken math, improving accuracy and interactivity in AI-driven educational tools.
- Kyutai released Voice of Reason, two speech-to-speech models aimed at solving math problems out loud using reinforcement learning.
- The models improved accuracy on spoken GSM8K math problems from 27.3% to 77.1%, utilizing no transcription step or separate text model.
- Voice of Reason leverages 150,616 math problems for supervised fine-tuning and runs on accessible checkpoints for self-hosting.
- The speech-native models offer interactive performance while maintaining naturalness in response time, significantly reducing latency compared to traditional cascaded systems.
Kyutai focuses on advancing artificial intelligence through novel speech processing methods. Prior research had limitations in accuracy, which Voice of Reason addresses through reinforcement learning. Future developments may enhance capabilities in educational AI applications and other speech-related tasks.
Read the original ↗
Nokia Open-Sources AnyJev: A Training-Free Layer That Turns Any Open LLM...
Nokia's open-source AnyJev library enhances the functionality of open large language models (LLMs) by enabling them to work as calibrated decision models, which can improve automated decision-making processes significantly.
- Nokia has open-sourced AnyJev, a Python library that converts open LLMs into decision models without the need for training.
- AnyJev utilizes a mechanism that corrects for biases in decision-making related to position and label preference.
- The library can be deployed from PyPI and supports various backends, including transformers and vLLM.
- On specific test tasks, the accuracy of decision-making improved, demonstrating the library's practical benefits for automatable tasks.
AnyJev is designed to transform open large language models (LLMs) into decision models that handle multiple-choice and scoring questions. Previously, biases were common in decision-making models that could affect their accuracy, and Nokia's solution aims to address these issues. Future developments may reveal further improvements or wider applications of this technology in automated systems.
Read the original ↗
Palma Raises $1.8M to Build the Governance Layer for Enterprise AI Agents
As enterprises increasingly deploy AI agents that perform actions, the need for a governance layer to ensure security and compliance is growing, making this technology crucial for managing risks.
- Palma.ai has secured $1.8 million in pre-seed funding to develop governance solutions for enterprise AI agents.
- The funding round was led by D11Z and included investments from multiple VCs and angel investors.
- Palma.ai's platform aims to create a governance layer that facilitates secure access for AI agents to enterprise systems, mitigating security risks.
- The platform is designed to be interoperable across different AI applications, allowing it to support various organizational tools and systems.
Palma.ai focuses on developing governance infrastructure for AI agents as enterprises shift from content generation to automated actions. The funding will enhance Palma's engineering and market reach as it aims to address security concerns linked to AI agent operations. This aligns with growing industry needs for formalized governance standards.
Read the original ↗
‘We’re already fighting yesterday’s battle’: Greece’s prime minister gets candid about AI
The candid remarks by Greece's prime minister about AI indicate a broader global recognition of the challenges posed by rapidly advancing technology, emphasizing the urgency for governments to adapt and address these changes.
- Greek Prime Minister Kyriakos Mitsotakis acknowledged a lack of preparedness for the challenges posed by AI during a trade mission in San Francisco.
- Mitsotakis described Greece's economic improvements, noting that it borrows more cheaply than the U.S. at a significant change from previous financial crises.
- Greece plans to implement a ban on social media for children under 15, which Mitsotakis acknowledges may already be outdated in the context of AI advancements.
- Greece is attracting investment in AI infrastructure, including data centers from Microsoft and AWS, without significant public opposition.
- The prime minister emphasized the need for 'smart regulation' for AI, reflecting concerns about job displacement and the implications of AI exceeding human capacities.
Kyriakos Mitsotakis is the Prime Minister of Greece, promoting the country's investment appeal while addressing the challenges of emerging technologies like AI. Greece has recently invested in digital infrastructure and is regaining developed market status. The country is becoming a focal point for dialogue and investment in technology and AI-related projects.
Read the original ↗
Trump Says AI Is Now Super Intelligence. What’s Really Super Is the Internet Response
Our editor hasn't briefed this story yet.
Read the original ↗
TechCrunch Founder Summit’s agenda revealed: Unlock fundraising, hiring, and AI insights in Boston on November 4
The TechCrunch Founder Summit on November 4 will provide valuable insights into startup strategies, giving founders access to expert guidance that can impact their business outcomes significantly.
- The TechCrunch Founder Summit is scheduled for November 4 at SoWa Power Station in Boston.
- The summit will cover crucial startup topics including fundraising, hiring, and AI strategies with experienced speakers.
- Attendees will gain insights directly from industry leaders on avoiding common pitfalls and optimizing company growth.
TechCrunch is a technology-focused media company that hosts events to support startups. The Founder Summit aims to educate entrepreneurs on key challenges they face in building successful businesses. The event will feature various sessions led by experienced investors and founders offering practical advice.
Read the original ↗
Qualcomm launches two new smartphone chips with emphasis on AI
These advancements in Qualcomm's smartphone chips could significantly enhance AI capabilities in consumer devices, impacting how users interact with technology on a daily basis.
- Qualcomm launched two new smartphone processors, Snapdragon 8 Elite Gen 6 and Snapdragon 8 Elite Extreme Gen 6, focused on AI features.
- The Extreme version can run a 30-billion-parameter mixture-of-experts model locally, improving AI processing efficiency.
- Both chips enable advanced voice features, including noise reduction and vocal boosting, enhancing call quality.
- Motorola announced the Motorola Signature 27 smartphone, which will feature the Snapdragon 8 Elite Extreme Gen 6 chip this year.
Qualcomm is a leading semiconductor company known for its mobile processors, particularly in smartphones. The new chips aim to facilitate more advanced AI features amid increasing competition in the tech market, particularly from companies like Apple. Future developments should focus on how these chips integrate with various high-end smartphones and applications.
Read the original ↗
Snorkel AI triples valuation to $3.5B as demand for AI training data booms
Snorkel AI's tripling valuation reflects a significant increase in demand for AI training data, highlighting the rapid growth of the data-as-a-service market in artificial intelligence.
- Snorkel AI raised $350 million in Series E funding, increasing its valuation to $3.5 billion.
- The startup had a previous valuation of $1.3 billion following a Series D funding round 17 months ago.
- Annualized revenue has grown to $375 million, an increase of eighteenfold over the past year.
- Snorkel's shift to data-as-a-service allows it to provide clients with completed datasets rather than just software for data-labeling automation.
- The funding round was led by Insight Partners and S32, with existing investors also participating.
Snorkel AI helps AI labs and companies build training datasets and simulate environments. Launched commercially in 2019, it focuses on providing completed datasets using a hybrid of synthetic data generation and human expertise. The competitive landscape includes other data companies seeing rapid growth in the AI sector.
Read the original ↗