01 ChatGPT turns product search into a shopping surface
Learn how to build, use, and scale workspace agents in ChatGPT to automate repeatable workflows, connect tools, and streamline team operations.
Related coverage also includes Introducing workspace agents in ChatGPT / OpenAI now lets teams make custom bots that can do work on their own, making this look more like a broader theme than a stray post.
The cluster is corroborated across 2 sources, so it reads as a developing trend rather than a one-off mention.
- Workspace agents are cloud‑hosted, Codex‑driven bots for automating repeatable team workflows.
- OpenAI intends the feature for paid plans including Business, Enterprise, Edu, and Teachers.
- Agents can connect to team tools (for example, Slack) to run tasks like web research, reporting, and parts of sales processes.
02 OpenAI opens GPT‑5.5 Bio Bug Bounty to crowdsource red‑teaming
What happened and why it matters: OpenAI launched a GPT‑5.5 Bio Bug Bounty, a red‑teaming challenge that asks researchers to find universal jailbreaks or other ways the model could produce bio‑risk content. The program offers monetary rewards, drawing external security researchers into model safety testing.
Scope and rewards: According to OpenAI’s announcement, the bounty focuses specifically on bio‑safety risks and offers rewards up to $25,000 for valid submissions. The challenge is framed as a proactive step to surface safety vulnerabilities before misuse can occur.
Intended outcome and context: By crowd‑sourcing red‑teaming, OpenAI aims to discover robust, transferable jailbreaks and mitigate risks in GPT‑5.5. This follows broader industry practice of using external researchers to test models under adversarial conditions.
- The bug bounty specifically targets GPT‑5.5 bio‑safety vulnerabilities.
- Rewards are offered up to $25,000 for qualifying findings.
- OpenAI is using external red‑team contributions to surface universal jailbreaks and reduce safety risks.
03 White House warns of 'industrial‑scale' efforts in China to copy U.S. AI tech
What happened and why it matters: U.S. officials warned that there are organized, large‑scale efforts in China to replicate or appropriate American AI technology, raising national‑security and industrial‑policy concerns. The warning frames AI competitiveness and intellectual property as a strategic issue for U.S. policy.
Reporting and context: Coverage cites senior U.S. sources describing systematic activities to acquire or reverse‑engineer U.S. AI advances. The concern is that rapid copying at scale could accelerate foreign deployments of technologies developed in the United States and complicate export‑control and trade responses.
Policy implications: The warning signals potential escalations in diplomatic and regulatory measures to protect AI-related IP and supply chains, and it underscores why governments are focusing on both technology governance and international enforcement to address AI proliferation risks.
- U.S. officials described the activity as organized, “industrial‑scale” efforts to copy U.S. AI technology.
- The warning links AI IP and competitiveness to national‑security policy responses.
- Officials’ statements increase pressure on export‑control, trade, and enforcement measures addressing AI supply chains.
What moved around the edges
Microsoft adds Agent Mode to Office while Anthropic wires Claude into personal apps
Microsoft is rolling out an Agent Mode inside Word, Excel, and PowerPoint this week — a more powerful Copilot experience the company has called "vibe working." Anthropic added connectors that let Claude access personal services such as Spotify, Uber Eats, Audible, Instacart and TurboTax.
The Verge AIGoogle puts Gemini 'auto browse' into Chrome for enterprises and launches an IT‑focused agent platform
Google added Gemini‑powered "auto browse" capabilities to Chrome for enterprise users to automate tasks like research and data entry. TechCrunch also reports the Gemini Enterprise Agent Platform is designed primarily for IT and technical users.
TechCrunch AIOpenAI publishes an open‑weight Privacy Filter to detect and redact PII
OpenAI released Privacy Filter as an open‑weight model intended to detect and redact personally identifiable information in text, with the company positioning it as state‑of‑the‑art for PII protection.
OpenAI BlogOpenAI unveils GPT‑5.5, billed as faster and more capable for coding and research
OpenAI introduced GPT‑5.5 and described it as faster and more capable than prior releases, built to handle complex tasks such as coding, research, and data analysis across tools.
OpenAI BlogOpenAI Blogで開発ツールの新提案
A deep dive into the Codex agent loop, showing how WebSockets and connection-scoped caching reduced API overhead and improved model latency.
OpenAI Blog