Welcome to P3 Media’s AI Commerce Brief, your daily update on the AI and commerce stories shaping how companies build, sell, and grow. It’s Wednesday, September 30. Let’s get into it.
OpenAI used its DevDay conference to push beyond the chatbot and toward always-on digital labor.
The company introduced Dots, agents that can keep working in the background on ongoing projects. Each dot gets its own cloud computer and can connect with outside apps through plugins.
Users can inspect the agent’s work, set permissions, and require approval for certain actions.
OpenAI says Dots are beginning to roll out to Pro, Business Premium, and Enterprise users in eligible markets.
That matters because the competitive question is shifting.
Model quality still matters, but distribution, permissions, and persistent work are becoming the product.
For business leaders, the practical test will be whether an agent can handle real workflows without creating unacceptable security or review burdens.
OpenAI also released GPT-6.1 Sol.
The company says the model improves on GPT-6 Sol across coding, document work, computer use, and multistep business workflows.
Standard API pricing is $2 per million input tokens and $10 per million output tokens.
OpenAI says that gives developers capabilities approaching its higher-end Astra model at a substantially lower cost.
The model is available in ChatGPT Work, Codex, and the OpenAI API, but not yet in regular Chat.
Meta is making a parallel move into business operations.
The company announced Muse for Small Business, adding skills and connectors for tools including Shopify, Stripe, QuickBooks, Klaviyo, Canva, Slack, and Meta’s own business accounts.
Meta says Muse can analyze business activity, draft campaigns, flag work that needs attention, and help with financial and operational reviews.
The company also says nothing publishes, sends, or spends without the user’s approval.
For merchants and agencies, the strategic point is bigger than another assistant.
Meta is positioning an agent close to the storefront, the ad account, the customer record, and the payment stack.
That could make the interface between commerce software and advertising platforms more conversational, but businesses will still need clear rules for access, data use, and approvals.
Anthropic released Claude Sonnet 5.5.
The company describes it as a faster, lower-cost complement to Claude Opus 5.5 for everyday work such as coding and document creation.
Anthropic says it runs more than 30 percent faster than Sonnet 5 and can cost up to 30 percent less per task, even though its token prices remain $2 for input and $10 for output per million tokens.
Sonnet 5.5 is available through Anthropic, Amazon Web Services, Google Cloud, and Microsoft Azure.
Those performance and efficiency figures are company claims, so buyers should validate them on their own workloads.
NVIDIA, meanwhile, launched the Open Agent Safety Platform.
It combines open-source OpenShell software with a Sentry reference design that uses BlueField-4 data processing units.
NVIDIA says OpenShell creates a secure runtime boundary around agents, while Sentry can monitor behavior independently and quarantine an agent that moves outside its boundaries.
The broader signal is that agent security is becoming an infrastructure category, not just an application feature.
That theme also reached Washington.
President Trump and leaders from Anthropic, Google, Meta, OpenAI, NVIDIA, and xAI signed a voluntary AI accord.
The document calls for internal controls, outside audits, board-level oversight, and monitoring designed to keep advanced systems from accessing technical systems in unintended ways.
Reuters and the Associated Press report that the accord leaves the door open to future regulation, but it is not itself a law.
What should operators watch next? First, whether persistent agents earn enough trust to become part of daily operating systems. Second, whether lower model costs translate into lower total workflow costs once supervision and security are included. Third, whether voluntary safety commitments become enforceable standards.
That’s your AI Commerce Brief for today. Thanks for listening.
OpenAI used its DevDay conference to push beyond the chatbot and toward always-on digital labor.
The company introduced Dots, agents that can keep working in the background on ongoing projects. Each dot gets its own cloud computer and can connect with outside apps through plugins.
Users can inspect the agent’s work, set permissions, and require approval for certain actions.
OpenAI says Dots are beginning to roll out to Pro, Business Premium, and Enterprise users in eligible markets.
That matters because the competitive question is shifting.
Model quality still matters, but distribution, permissions, and persistent work are becoming the product.
For business leaders, the practical test will be whether an agent can handle real workflows without creating unacceptable security or review burdens.
OpenAI also released GPT-6.1 Sol.
The company says the model improves on GPT-6 Sol across coding, document work, computer use, and multistep business workflows.
Standard API pricing is $2 per million input tokens and $10 per million output tokens.
OpenAI says that gives developers capabilities approaching its higher-end Astra model at a substantially lower cost.
The model is available in ChatGPT Work, Codex, and the OpenAI API, but not yet in regular Chat.
Meta is making a parallel move into business operations.
The company announced Muse for Small Business, adding skills and connectors for tools including Shopify, Stripe, QuickBooks, Klaviyo, Canva, Slack, and Meta’s own business accounts.
Meta says Muse can analyze business activity, draft campaigns, flag work that needs attention, and help with financial and operational reviews.
The company also says nothing publishes, sends, or spends without the user’s approval.
For merchants and agencies, the strategic point is bigger than another assistant.
Meta is positioning an agent close to the storefront, the ad account, the customer record, and the payment stack.
That could make the interface between commerce software and advertising platforms more conversational, but businesses will still need clear rules for access, data use, and approvals.
Anthropic released Claude Sonnet 5.5.
The company describes it as a faster, lower-cost complement to Claude Opus 5.5 for everyday work such as coding and document creation.
Anthropic says it runs more than 30 percent faster than Sonnet 5 and can cost up to 30 percent less per task, even though its token prices remain $2 for input and $10 for output per million tokens.
Sonnet 5.5 is available through Anthropic, Amazon Web Services, Google Cloud, and Microsoft Azure.
Those performance and efficiency figures are company claims, so buyers should validate them on their own workloads.
NVIDIA, meanwhile, launched the Open Agent Safety Platform.
It combines open-source OpenShell software with a Sentry reference design that uses BlueField-4 data processing units.
NVIDIA says OpenShell creates a secure runtime boundary around agents, while Sentry can monitor behavior independently and quarantine an agent that moves outside its boundaries.
The broader signal is that agent security is becoming an infrastructure category, not just an application feature.
That theme also reached Washington.
President Trump and leaders from Anthropic, Google, Meta, OpenAI, NVIDIA, and xAI signed a voluntary AI accord.
The document calls for internal controls, outside audits, board-level oversight, and monitoring designed to keep advanced systems from accessing technical systems in unintended ways.
Reuters and the Associated Press report that the accord leaves the door open to future regulation, but it is not itself a law.
What should operators watch next? First, whether persistent agents earn enough trust to become part of daily operating systems. Second, whether lower model costs translate into lower total workflow costs once supervision and security are included. Third, whether voluntary safety commitments become enforceable standards.
That’s your AI Commerce Brief for today. Thanks for listening.
