OpenAI Launches Dots AI Agents for ChatGPT Subscribers

Key Takeaways

• <ul><li>OpenAI launched Dots AI agents and the affordable GPT-6.1 Sol model at Developer Day</li><li>Dots agents compete directly with Meta's Muse and are aimed at enterprise users</li><li>OpenAI cancelled a more powerful Astra model release due to safety concerns</li><li>Google Gemini has structural advantages through Android integration and 82% Cloud revenue growth</li><li>Amazon invested in 700 renewable energy projects generating 42 gigawatts of power</li><li>HP launched AI-focused laptops targeting $58.3 billion in revenue by 2029</li><li>TrendAI expanded NVIDIA's Agent Safety Platform with threat intelligence capabilities</li><li>Saviynt appointed a new Japan leader to expand identity security for AI adoption</li><li>Researchers developed P-TTT to improve personalized AI reward models</li><li>Trump considers Jay Clayton as AI czar and created an AI security task force</li></ul>

OpenAI launches Dots AI agents to compete with Meta's Muse

OpenAI announced its new AI agents called Dots at its developer conference in San Francisco. These agents can perform tasks like booking reservations or managing finances without direct user input. They compete with Meta's agent called Muse. Dots will be available to ChatGPT Pro subscribers and business customers starting Tuesday, though some European users are excluded. Users can set rules for what agents can do and must approve sensitive actions.

OpenAI releases Dots agents and cheaper GPT-6.1 Sol model

OpenAI introduced Dots AI agents powered by its Astra model at Developer Day. The agents can handle tasks like scheduling and submitting invoices after getting user approval. OpenAI also launched GPT-6.1 Sol, a mid-range AI model priced at one-fifth the cost of its flagship GPT-6 Astra. Dots will roll out to Pro, Business Premium, and Enterprise users. The company is preparing for a potential IPO early next year.

OpenAI unveils affordable GPT-6.1 Sol and autonomous Dots agents

OpenAI launched GPT-6.1 Sol, a more affordable AI model, alongside autonomous agents called Dots. The announcement came amid growing competition with Anthropic and Meta. Just a day earlier, OpenAI cancelled the release of a more powerful Astra model due to safety concerns about misleading behavior. Dots are designed for the enterprise market and can work across multiple applications. OpenAI is investing heavily in AI infrastructure and eyeing a potential IPO valued over $1 trillion.

Trump considers Jay Clayton as AI czar

President Trump is considering naming Jay Clayton as the administration's AI czar. Clayton is a former Goldman Sachs executive and close friend of the president. He has been working on developing a national AI strategy. The position has been vacant since Andrew Ng departed. Sources say Trump is leaning toward the appointment, though no final decision has been made.

Trump administration creates AI security task force

The Trump administration established a new task force to oversee AI system development and deployment. The task force will ensure AI systems prioritize security and accountability. The administration is also requiring companies to disclose AI use in their products. These steps follow recent security breaches involving compromised AI systems. Some experts worry the measures may not fully address complex AI security issues.

New P-TTT Method Improves Personalized AI Rewards

Researchers found that current methods for personalized reward models fail to capture specific user preferences effectively. They created a new approach called Preference-Aligned Test-Time Training, or P-TTT, to fix this issue. This method updates fast weights quickly using a single forward pass without needing complex backpropagation. Tests show that P-TTT captures historical preference relations better than existing state-of-the-art methods.

TrendAI Adds Security to NVIDIA Agent Safety Platform

TrendAI has expanded the NVIDIA Agent Safety Platform by adding threat intelligence and full AI factory security. The new TrendAI Vision One system combines hardware detection on NVIDIA BlueField DPUs with network security features. It also includes Zero Day Initiative threat intelligence to help enterprises protect their AI agent fleets. This partnership aims to create a safer environment for moving AI systems from pilot programs to production use.

Google Gemini May Outperform Meta Muse in AI Race

Although Meta launched its agentic AI model Muse before Google, Google has a structural advantage that could help it win the market. Google is integrating Gemini directly into its operating system on Android, Maps, and Gmail, while Meta relies on chat apps. Google Cloud revenue grew 82% and nearly 90% of Fortune 100 companies now use Gemini Enterprise. Google also spent $44.92 billion on capital projects and raised debt to $98.2 billion to fund its AI expansion.

New AI Laptops Could Boost HP Stock Value

HP Inc. launched two new AI-focused laptops, the HP Googlebook 14 and the ZBook Ultra G3a mobile workstation. These devices are designed to run large local AI models directly on the hardware instead of relying on the cloud. Investors believe these high-specification PCs can help offset slow growth in HP's printing business. Analysts forecast HP revenue of US$58.3 billion and earnings of US$2.7 billion by 2029.

Climate Week NYC Highlights AI and Wildfire Tech

Leaders gathered at Climate Week NYC to discuss solutions for the environmental impact of AI data centers and climate change. Amazon executives announced investments in 700 renewable energy projects that generate 42 gigawatts of power. The event also celebrated youth innovation, featuring Satellites on Fire, a group founded by 16-year-olds in Argentina. Their AI system analyzes data from 13 satellites every five minutes to detect wildfires early and send alerts via WhatsApp.

AI shopping agents compete for your online purchases

Businesses are beginning to figure out how to handle AI agent visitors on their shopping sites. The article discusses how AI agents are competing to influence online shopping decisions. Major retailers like Amazon, Walmart, and Wayfair are involved. The situation is changing how e-commerce works.

Saviynt names Michibumi Serizawa to lead Japan AI security expansion

Saviynt appointed Michibumi Serizawa as Vice President and Country Manager for Japan. The company is expanding in Japan through a joint venture with Ashisuto. Serizawa has over 30 years of IT experience from companies like Dell, F5 Networks, Palo Alto Networks, and Wiz. Saviynt is focused on identity security as AI adoption grows in Asia Pacific and Japan. The company launched its new Japanese subsidiary to help enterprises deal with AI-era identity challenges.

Study finds deterministic AI guidance underperforms in reasoning tasks

Researchers studied discrete diffusion language models and found that deterministic process reward model guidance loses to simpler methods. On Dream-v0-Instruct-7B, deterministic guidance reached 65.18% while independent sampling with an outcome reward model reranker reached 75.13% on GSM8K problems. The gap grew to 12.69 percentage points with 32 candidates. The study identified two key failures in deterministic guidance and released tools for reproducible comparisons at matched compute budgets.

Sources

• OpenAI Unveils Dots, New A.I. Agents to Rival Meta’s Muse
• OpenAI debuts Dots AI agents in challenge to Meta's popular Muse agent
• OpenAI unveils cheaper GPT-6.1 Sol and autonomous Dots agents amid intensifying AI race
• Scoop: Trump warms to Jay Clayton as AI czar
• Trump Admin Addresses AI Security Concerns Around America.gov
• Using Context Is Not Enough: Test-Time Training for Personalized Reward Modeling
• TrendAI™ Extends the NVIDIA Agent Safety Platform with Threat Intelligence and Full AI Factory Security
• Muse Got Here First But This Agentic AI Will Supercede It Soon
• Does New AI Laptops Change The Bull Case For HP (HPQ)?
• From AI data centers to NASA-beating wildfire tech: 5 lessons from Climate Week NYC
• Big Tech’s AI agents are fighting over your shopping cart
• Saviynt Appoints Michibumi Serizawa to Lead Japan’s AI-Era Identity Security Growth
• Why Deterministic PRM Guidance Underperforms in Discrete Diffusion Reasoning