Cohere released Command A+, an open-weights mixture-of-experts model targeting enterprise agentic workloads with sovereign deployment as a core feature.

• Benchmark position: ranks first on Artificial Analysis non-hallucination at 86%, placing it near Claude 4.5 Haiku on the Intelligence Index at a fraction of the compute cost.
• Sovereign angle: open weights plus Cohere's on-premises deployment options make Command A+ the clearest answer to enterprise buyers who cannot route data through U.S. hyperscaler APIs.

Tenable announced general availability of Hexa AI, an agentic engine inside the Tenable One platform that executes multi-step remediation workflows with MCP support for custom agent building.

• MCP integration means security teams can wire Hexa into existing toolchains without replatforming, lowering the adoption barrier for agentic security automation.
• The GA timing puts Tenable directly in competition with CrowdStrike and SentinelOne as all three race to make AI the default remediation path, not just a detection aid.

Zscaler announced intent to acquire Symmetry Systems, whose access graph maps how human and non-human identities, applications, and data connect, to govern AI agent communication at scale.

• Gap it fills: most zero-trust architectures were built for human users; Symmetry's identity graph extends that model to the millions of agent-to-agent calls that existing tools cannot see.
• Strategic timing: the acquisition lands as enterprises are deploying agentic workloads without mature governance, making visibility into agent data access the new compliance frontier.

Cloudflare and Anthropic integrated Claude Managed Agents with Cloudflare Sandboxes, splitting the agent control loop from code execution so tool calls and network connections run on Cloudflare infrastructure rather than Anthropic's.

• Architecture: the Claude agent session lives on Anthropic's platform while a Cloudflare Workers control plane assigns execution environments, giving enterprises network-level isolation without managing their own compute.
• Enterprise relevance: separating inference from execution addresses the data egress and audit concerns that have slowed regulated industries from adopting hosted agent platforms.

Nvidia reported Q1 FY2027 revenue of $81.6B, up 85% year-over-year with net profit up 210%, driven by data center demand from hyperscalers and frontier AI firms running inference at scale.

• Blackwell systems are in their fastest product ramp in Nvidia's history; partner data centers exceeding 1GW of AI compute have multiplied, and Q2 guidance is $91B, another sequential record.
• Vera Rubin ships Q3 2026 with GPU-initiated direct storage access and a memory bill of materials that has surged 435% versus Grace Blackwell, reflecting HBM4 scarcity pricing.