Need to know
- Meta Muse Spark replaces Llama 4 on Ray-Ban and Oakley smart glasses starting today in the US and Canada.
- Palo Alto Networks and Deutsche Telekom launched Sovereign Cortex, bringing AI-driven SecOps with data sovereignty controls to European regulated industries.
- BeyondTrust was selected for Anthropic's Project Glasswing to help secure critical software infrastructure.
- Cursor is opening a London EMEA headquarters this summer and targeting 200 employees by year-end.
- Microsoft Foundry shipped production-grade runtime, tooling, and governance layers for enterprise AI agents.
New Releases
Meta is replacing Llama 4 with Muse Spark on most of its smart glasses lineup, marking the first public deployment from Meta Superintelligence Labs.
- Rollout scope: Ray-Ban Meta and Oakley Meta in the US and Canada go live today; Meta Ray-Ban Display is scheduled for this summer.
- Model lineage: Muse Spark is the first in the Muse series, which formally succeeds the Llama series for edge and wearable deployments.
Microsoft Foundry today added a production runtime, developer tooling, and governance controls specifically designed for enterprise AI agent deployments.
- What shipped: The update adds runtime infrastructure for executing agents at scale alongside governance primitives for policy enforcement and auditability.
- Enterprise signal: Foundry's addition of governance before broad adoption suggests Microsoft is prioritizing compliance readiness over raw capability in its enterprise agent pitch.
Palo Alto Networks and Deutsche Telekom today announced Sovereign Cortex with T Security, delivering the Cortex AI-driven SecOps platform to European regulated industries with data sovereignty controls independently governed by Deutsche Telekom.
- Sovereignty mechanism: Data governance is managed independently by Deutsche Telekom, not Palo Alto Networks, addressing the regulatory barrier that has blocked US-headquartered security platforms from EU financial and critical infrastructure sectors.
- Target market: The joint solution is aimed at financial services, energy, and other regulated European industries where DORA and NIS2 compliance requires data residency guarantees.
GitHub code scanning now automatically runs every 30 days on repositories that have had no pushes or pull requests for six months or more, closing a coverage gap that left dormant codebases unmonitored.
- Configuration: Organizations can enable the setting globally from their security settings; it applies to all repos using code scanning default setup, with no per-repo configuration required.
- Risk addressed: Legacy and archived codebases are disproportionately targeted in supply chain attacks because they accumulate known CVEs without triggering standard CI-based scans.
Funding
Cursor is formalizing its EMEA expansion with a London headquarters launching this summer and a plan to reach 200 employees by year-end, while separate reports indicate SpaceX is exploring a potential acquisition of the coding platform.
BeyondTrust has been selected for Project Glasswing, Anthropic's collaborative program to harden critical software infrastructure, placing it among a curated set of vendors whose codebases will be integrated with Anthropic's security research.
Case Studies
Notion disclosed that an internal workplace coordinator used its developer platform to automate new-hire office seat assignments, connecting Notion with OfficeSpace and Slack via custom agents and eliminating recurring manual manager coordination.
- Architecture used: The workflow uses a Notion Worker and custom agents to bridge Notion, OfficeSpace, and Slack without any bespoke middleware or third-party automation platform.
- Signal value: Notion eating its own cooking on agentic workflows gives enterprise buyers a concrete reference architecture before building their own.
Trending on X
- Anthropic vs OpenAI IPO race Prediction market odds on which company goes public first keep flipping, with Anthropic filing confidentially for June 9 and OpenAI having filed weeks earlier, and the community debating whether Anthropic's Mythos release timed alongside the filing is deliberate positioning.
- Anthropic Mythos public release timing Anticipation is high for Mythos going broadly available today after a restricted preview period, with debate centering on whether public access will expose the same 10,000-zero-day capability that alarmed financial regulators during the closed beta.
- AI labs on slowing AI development Ethan Mollick and others are reacting to both Anthropic and OpenAI including language about potentially slowing AI development in their roadmap posts, while noting neither has identified a practical coordination mechanism to make that happen.
- Apple WWDC platform risk for AI labs Practitioners are debating whether Apple, Microsoft, and Google now control the user-facing AI layer so completely that OpenAI and Anthropic are structurally disadvantaged regardless of model quality.
- AI agents breaking Zero Trust models Security architects are discussing Okta's framing that autonomous agents operate outside the behavioral assumptions baked into Zero Trust, with identity provenance and short-lived task-scoped credentials emerging as the proposed mitigation.