Anthropic introduces a proactive AI agent that works across connected apps, while Google expands Gemini, Apple unveils new AI-powered hardware, and competition over enterprise customers grows.
Anthropic has introduced Claude Duet, a proactive artificial intelligence assistant designed to manage tasks, monitor connected applications and continue working when users are away. The company announced the research preview on October 21, 2026, as competition among major technology companies increasingly shifts toward AI systems that can perform work independently.
The new assistant can review emails, manage calendars, prepare documents and identify tasks requiring attention. Unlike traditional chatbots that respond primarily to direct questions, Claude Duet can initiate contact when it finds something important and continue handling assignments without constant supervision.
Anthropic initially made the service available through a waitlist for eligible paid subscribers in the United States. The launch comes during a busy period for the AI industry, with Google releasing a new flagship model, Apple introducing advanced hardware and technology companies facing growing questions about automated decision-making.
How Claude Duet Works
Claude Duet operates through a cloud-based virtual computer that allows it to continue completing assignments after a user closes the application. It can access approved services and perform tasks across multiple applications without requiring the user’s computer or smartphone to remain active.
Anthropic designed the system for professionals, entrepreneurs and small-business owners who routinely manage competing responsibilities. The assistant can organize information from connected accounts, prepare daily briefings and alert users when important developments require attention.
Each morning, Claude Duet can produce a summary of significant information collected from connected applications. Throughout the day, it periodically checks for new developments and sends notifications when necessary.
The assistant also supports more complicated work, including creating financial forecasts, developing dashboards, preparing reports and building web applications. Users can communicate with it through written messages or voice instructions.
Claude Duet uses Anthropic’s Claude Opus 5.5 model and integrates with the company’s existing system of connected applications. The service can work with tools such as Gmail, Slack, QuickBooks, Stripe and other business platforms.
Anthropic says the assistant can use more than 6,000 connectors and plugins available through its marketplace. However, the usefulness of those integrations depends on the connected services, permissions and specific tasks involved.
Human Approval Remains Part of the Process
Although Claude Duet can operate independently, Anthropic says it requires user approval before taking certain significant actions. These include sending messages, making purchases, signing documents and deleting information.
The company has also introduced safeguards intended to prevent sensitive information from being shared without authorization. Its security systems examine proposed actions and attempt to detect malicious instructions hidden in emails, websites and other external material.
These measures address a major concern surrounding autonomous AI agents. Systems that access private correspondence, financial accounts and business records may encounter information they should not disclose or instructions they should not follow.
Anthropic acknowledged that Claude Duet remains an experimental product. The assistant can misunderstand conversations, misjudge priorities or make mistakes when completing complicated assignments.
The company also explained that background activity consumes subscription usage limits. Users who assign frequent monitoring tasks or complex projects may reach those limits faster than people using a conventional chatbot.
Claude Duet is available as a separate application for iOS and Android and through Anthropic’s web platform. Its initial research preview is restricted to users admitted from the waitlist rather than being available to everyone immediately.
Google Expands Gemini 4 Argon
Anthropic’s announcement follows Google’s October 13 expansion of Gemini 4 Argon, its advanced artificial intelligence model designed for demanding professional and technical assignments. The system is available to developers and businesses through Google’s AI platforms and to eligible subscribers through consumer products.
Google says Gemini 4 Argon is designed to handle complicated problems involving software development, financial analysis, legal work, cybersecurity and scientific research. The model can maintain context across lengthy assignments and coordinate multiple steps toward completing a task.
The company also provides access through Google Cloud, allowing organizations to incorporate the technology into their existing systems. Reuters reported that Google has been promoting the model to business customers as competition for enterprise AI spending intensifies.
Google’s introductory application programming interface pricing starts at $2 per million input tokens and $10 per million output tokens. Those rates are scheduled to increase after November 30, according to the company’s published pricing information.
The competition reflects a broader change in the industry. Technology companies increasingly want their systems to perform useful work across software platforms rather than simply generate text, answer questions or summarize documents.
Anthropic Prepares for a Public Offering
Anthropic is also pursuing a major financial milestone. On October 26, the company announced that it had filed a registration statement with the U.S. Securities and Exchange Commission for a proposed initial public offering.
The company intends to list its Class A common stock on the Nasdaq Global Select Market under the ticker symbol ANT. However, the number of shares being offered and their proposed price range have not yet been announced.
The filing does not mean Anthropic has completed an initial public offering. Its registration statement had not become effective when the company announced the proposal, leaving the timing and final terms unresolved.
In May, Anthropic announced that it had raised $65 billion through Series H financing. The investment valued the company at approximately $965 billion after the funding round.
Anthropic also reported that its annualized revenue run rate had exceeded $47 billion during May. That figure represents revenue projected over a year based on a recent operating pace, not necessarily revenue already earned during a completed financial year.
The enormous funding round demonstrates investors’ expectations for continued growth in commercial AI services. It also illustrates the financial demands of developing advanced models and operating the computing infrastructure required to support them.
OpenAI and Anthropic Compete for Business Customers
OpenAI and Anthropic are competing to secure large business customers, whose contracts can involve substantial spending on AI services. Pricing, available computing capacity, model performance and contractual flexibility all influence those decisions.
A September report from The Information described tensions surrounding Anthropic’s enterprise discount policies. According to the report, some customers faced higher charges after exceeding discounted usage amounts, creating opportunities for competitors offering different arrangements.
The original claim that OpenAI was providing discounts of up to 90% specifically to attract Anthropic customers could not be independently verified from reliable documentation reviewed for this article. Negotiated enterprise discounts should not be confused with publicly available subscription pricing.
Anthropic’s proposed stock listing adds another dimension to the rivalry. Its financial disclosures and future performance could offer investors more information about the economics of advanced AI development and commercial adoption.
Apple Introduces New AI-Powered MacBooks
Apple has also entered the latest round of technology announcements with a redesigned MacBook Pro featuring an OLED display and touchscreen capabilities. The company unveiled the new computers on October 28, bringing touch interaction to the MacBook Pro for the first time.
The updated laptops feature Apple’s M5 Pro and M5 Max processors, redesigned displays and additional capabilities connected to Apple Intelligence. Apple began accepting preorders on October 28, with general availability scheduled for November 6.
The new hardware demonstrates Apple’s approach to combining AI features with its existing products. Rather than focusing exclusively on standalone AI agents, Apple is integrating intelligent functions into its operating systems, devices and established applications.
Apple chief executive John Ternus discussed the company’s touchscreen strategy in an interview with Wired. He explained that Apple wanted to introduce touch without compromising the MacBook’s traditional keyboard-and-trackpad experience.
The announcement follows Apple’s continuing efforts to improve Siri and expand its artificial intelligence features. The company’s emphasis remains on making those capabilities accessible through familiar consumer devices.
AI Safety Faces Greater Public Scrutiny
As automated systems become more capable, government officials are examining whether existing safety measures provide adequate protection. On October 5, the New York City Council questioned representatives from OpenAI, Anthropic, Google and Meta during a hearing about artificial intelligence risks.
The Associated Press reported that former employees of major AI companies warned lawmakers about the pace of development and the difficulty of ensuring that increasingly powerful systems remain under human control. Current company representatives defended their safety efforts while expressing varying degrees of openness to regulation.
Council members examined questions involving transparency, independent oversight and accountability for harmful AI behavior. The hearing also considered proposed safeguards intended to protect consumers and encourage employees to report serious safety concerns.
These issues become especially important when AI agents gain access to private accounts and receive permission to act on behalf of individuals or businesses. A mistake involving an email summary can be inconvenient, while an incorrect financial transaction or disclosure of confidential information may have more serious consequences.
Claude Duet’s approval requirements and security controls are intended to reduce those risks. Their effectiveness across different applications and unpredictable real-world situations will require further testing as access expands.
Anthropic’s next steps include admitting additional users to the Claude Duet research preview and continuing development based on their experiences. The company is also pursuing its proposed Nasdaq offering, while Google, Apple and other competitors expand their AI products during the final months of 2026.
IMAGE: Source Own work Author 123net CC3