日本語
2026-10-22 Morning edition
Morning EditionAI News Daily
2026-10-22
Always-on personal AI agents are heating up, alongside a wave of new safety and privacy features.
2026-10-22 Morning edition
Today's Highlights
FeatureAnthropic launches research preview of always-on personal agent Claude Duet
2026-10-21
CompanyNVIDIA unveils OpenShell, a reference architecture for safe AI agents
2026-10-21
FeatureChatGPT adds Locked Chats, letting users hide conversations behind a PIN
2026-10-20
2026-10-22 Morning edition
Anthropic Opens Research Preview of Claude Duet
Published: 2026-10-21
- Built on Claude Opus 5.5, Duet is an always-on personal agent that proactively handles tasks across email, calendars, and 6,000+ integrations.
- It always confirms before sensitive actions like sending messages or payments, and is rolling out via waitlist to US Pro/Max subscribers.
Why it matters: Anthropic joins OpenAI's Dots and Meta's Muse in the always-on agent race, signaling that personal-task automation is now a three-way contest worth comparing closely.
KEY CONTEXTOpenAI, Meta, and Anthropic have each shipped an always-on agent, kicking off a head-to-head comparison phase.
2026-10-22 Morning edition
NVIDIA Unveils OpenShell Reference Architecture
Published: 2026-10-21
- At GTC Berlin, NVIDIA introduced OpenShell, a reference architecture for building safe AI agents.
- It standardizes permission management and oversight mechanisms to help enterprises deploy agents safely.
Why it matters: A shared safety foundation for agentic AI gives enterprises a concrete benchmark for managing risk in production deployments.
2026-10-22 Morning edition
ChatGPT Adds Locked Chats Feature
Published: 2026-10-20
- Locked Chats lets users protect specific conversations with a PIN or device authentication and hide them from regular history.
- It is rolling out across Free, Go, Plus, and Pro plans on web, iOS, Android, and the new desktop app.
- Resetting the PIN permanently deletes any locked chats, so users should note this before resetting.
Why it matters: This strengthens protection for sensitive information on shared devices or during screen sharing, giving business users more privacy control.
2026-10-22 Morning edition
Trend Overview
- Competition among always-on personal agents (Anthropic Duet, OpenAI Dots, Meta Muse) is intensifying.
- OpenAI is accelerating its business on multiple fronts, including cloud partnerships, M&A, and ad expansion.
- Building safety and oversight for agentic AI is emerging as a shared industry challenge.
2026-10-22 Morning edition
Summary
Top 3 Takeaways
- Anthropic launched always-on agent Claude Duet via waitlist.
- NVIDIA published OpenShell, a reference architecture for safe agents.
- OpenAI added PIN-protected Locked Chats to ChatGPT.
What to Watch Next
- Pricing and availability expansion for always-on agents across providers
- Whether OpenAI revises pricing or terms of service (unconfirmed as of this edition)
- Further updates and independent confirmation on NVIDIA Vera Rubin Infrastructure