OpenAI launched GPT-6 Astra today, its first model to cross the "Critical" cybersecurity threshold under its Preparedness Framework, triggering additional deployment restrictions.

• Enterprise admins must manually enable Astra for their workspace; initial access is limited to Daybreak cybersecurity customers, with ChatGPT Plus, Pro, Business, and Enterprise rollout following over the next few days via OpenAI API and AWS.
• Greg Brockman declared the "AGI era" has arrived, but Polymarket's best-model market for September still puts Anthropic at 71.5% versus OpenAI at 24.15%, suggesting the community is not yet convinced.

Salesforce and Anthropic announced Claudeforce today, an expanded strategic partnership that positions Claude as the default reasoning layer inside Agentforce and Slack.

• Claudeforce gives Claude agents direct access to Salesforce Data Cloud, letting them act on CRM context without data leaving the Salesforce security perimeter.
• The deal deepens a pattern of Anthropic embedding into enterprise platforms rather than selling direct, following the AWS and Google Cloud arrangements.

CrowdStrike today introduced an Agentic Identity Provider inside Falcon Next-Gen Identity Security, designed to issue verifiable identities and continuously authorize autonomous AI agents.

• Traditional identity providers assume human users with session-based auth; the Agentic IdP issues scoped credentials to agents, enforces least-privilege dynamically, and integrates with Falcon's threat detection for real-time revocation.
• The announcement came alongside the Rubrik-CrowdStrike agentic identity resilience workflow, where Charlotte Agentic SOAR orchestrates detection, investigation, and recovery from compromised agent identities end to end.

Google today moved Gmail Live, Docs Live, and Keep Live from beta to general availability for paying Google AI subscribers.

• Gmail Live lets subscribers search email by voice; Docs Live drafts documents from Drive and Chat context; Keep Live converts spoken prompts directly into notes, lists, and reminders on mobile.
• Access tiers vary by product and OS, meaning enterprise Workspace admins will need to map which features land for which subscription levels before communicating availability to employees.

Zscaler reported Q4 FY2026 revenue of $898M, up 25% year over year, beating the $877M consensus, with CEO Jay Chaudhry attributing momentum to urgent enterprise demand driven by AI-enabled threats.

• EPS came in at $1.19 adjusted versus $1.09 expected; net-new ARR excluding the Red Canary acquisition grew 17% in Q4, though FY2027 ARR guidance implies the pace slows to roughly 4% net-new ARR growth for the full year.
• Chaudhry cited agentic Zero Trust as an early-momentum product, the first public signal that zero-trust vendors are seeing commercial traction specifically from customers securing AI agent traffic, not just human users.

SentinelOne expanded Wayfinder Frontier AI Services with OpenAI Daybreak models today, after its SentinelLABS research arm benchmarked GPT-5.6-Cyber as best in class on military-grade malware reverse engineering.

• New capabilities include code risk analysis, compromise assessment, and malware analysis delivered through the Daybreak Defense Network, all starting with GPT-5.6-Cyber before expanding to GPT-6 Astra cyber models.
• The internal benchmark result, not just a commercial arrangement, drove the model selection decision, giving SentinelOne's procurement choice more credibility than a typical partnership announcement.