🤖 The Blockbrain Brief #31

AI is moving from experimental tools into persistent, secure, and enterprise-ready systems built for real workplace adoption.

📩 High-level Summary

• AI Assistants: OpenAI’s ChatGPT memory upgrades, Codex settings, and Ona acquisition show that AI tools are moving toward long-running, personalized work companions that can remember context and continue tasks securely.
• Enterprise Agents Advancements: Claude Managed Agents, Microsoft 365 Copilot updates, and DXC’s Claude alliance point to AI agents being embedded deeper into real business workflows, from scheduled reports to regulated enterprise systems.
• Adoption Is Now a Business Priority: Companies like Mendo and Zaro show that the next challenge is not just building AI tools, but helping employees actually use them effectively through better workflows, shared context, analytics, and onboarding.
• Trust and Transparency: Anthropic’s Claude Fable 5 backlash and Microsoft’s governance updates highlight that enterprises need AI systems to be safe, auditable, and transparent when limits or safeguards are applied.
• Europe Is Pushing for Sovereign AI: Mistral’s sovereign AI strategy and the EU Cloud and AI Development Act show that Europe is turning AI infrastructure, cloud control, and data jurisdiction into major enterprise and policy priorities.
• EU AI Infrastructure Investment: Databricks’ potential $175B valuation and Europe’s €2.3B tech funding week show continued investor confidence in AI, data infrastructure, enterprise software, and the tools powering AI adoption.

OpenAI Improves Codex Experience

OpenAI is continuing to improve the day-to-day experience of using Codex and ChatGPT. In Codex, new app settings give users more control over behavior, preferences, file-opening, command output, notifications, appearance, and profile activity insights such as lifetime tokens and peak token usage.

Should we put black holes to help warn clients of their usage? (just kidding)

Microsoft 365 Copilot Adds New Models, Notebooks, Skills, and Admin Tools

Microsoft has rolled out a broad set of Microsoft 365 Copilot updates, led by the addition of Anthropic’s Claude Opus 4.8 and OpenAI’s GPT-5.5 Instant. The updates also include a refreshed Copilot app design, stronger Copilot Notebooks capabilities, Teams meeting sources, Excel generation, infographics, and new built-in skills across Word, PowerPoint, and SharePoint.

On the enterprise side, Microsoft is pushing Copilot deeper into IT governance and adoption. It expanded ISO 42001 certification across its Copilot portfolio, launched the Copilot Adoption Hub, introduced federated MCP-based connectors for real-time enterprise data, and added watermarking for sensitive meeting content.

Claude Managed Agents Can Now Run on Schedules and Use Secure Vaults

Anthropic introduced two major public beta features for Claude Managed Agents: scheduled runs and vault-based environment variables. Scheduled deployments allow agents to automatically handle recurring tasks such as data syncs, compliance scans, reports, digests, log monitoring, and follow-ups without teams having to build their own scheduler.

The vault update is equally important for enterprise use. Agents can now securely access authenticated tools through environment variables stored in vaults, with real keys attached only at approved network boundaries. This allows agents to work with CLIs, MCP, APIs, and tools like Browserbase, KERNEL, Notion, Ramp, and Sentry without directly exposing sensitive credentials.

OpenAI Acquires Ona to Give Codex Persistent Cloud Execution

OpenAI announced plans to acquire Ona, a company focused on secure cloud execution and orchestration. The acquisition is meant to expand Codex with customer-controlled cloud infrastructure, allowing AI agents to work on complex, long-running tasks in secure environments even when a user is offline.

Instead of being limited to a local machine or active session, Codex agents could eventually continue work inside an organization’s cloud environment with stronger control over data, security boundaries, credentials, logging, and review workflows. OpenAI framed this as part of Codex’s evolution from a coding assistant into a broader system for software and knowledge work.

Anthropic and DXC Bring Claude Into Regulated Enterprise Systems

Anthropic and DXC Technology have formed a multi-year global alliance to embed Claude into mission-critical systems used by banks, airlines, insurers, manufacturers, and government agencies. DXC plans to train tens of thousands of Claude-certified forward-deployed engineers who will work directly inside customer organizations.

DXC first tested Claude internally, including in OASIS, its AI-native managed services platform. Claude helped generate more than 95% of the code for OASIS, which now serves over 50 customers. The partnership will begin across insurance, modernization as a service, cybersecurity, and application services. The main message is clear: Anthropic is not only selling model access, but also building a services ecosystem to help large enterprises actually deploy Claude in regulated environments.

Mendo Raises €12M to Help Enterprises Adopt Generative and Agentic AI

Paris-based Mendo raised €12 million in Series A funding to scale its platform for enterprise AI adoption. The company helps organizations identify useful AI cases, integrate AI into daily workflows, support employees, and turn AI investments into measurable outcomes. Its early results reportedly show adoption rates up to six times higher than traditional approaches.

The funding will support product development, especially analytics, as well as team growth from 50 to 100 employees and commercial expansion across Europe. Mendo’s core insight is that the bottleneck in enterprise AI is no longer just technology. Companies also need adoption strategy, workflow integration, measurement, and change management.

Zaro Emerges From Stealth With $5.1M to Build an Enterprise AI Context Layer

London-based Zaro has emerged from stealth with $5.1 million in pre-seed funding led by Cherry Ventures. The startup is building an AI-native workspace that connects enterprise data, workflows, decisions, and AI tools through a shared context layer. Its goal is to prevent organizational knowledge from getting trapped across disconnected agents, apps, and workflows.

Zaro’s platform combines shared context, application-building tools, and pre-configured workflows. It also uses a multi-model approach, routing simpler tasks to cheaper models while reserving advanced models for more complex work. This reflects a growing enterprise AI theme: the long-term value may not only come from the model itself, but from the company-specific context that models can use and reuse over time.

Anthropic Walks Back Hidden Claude Fable 5 Capability Limits

Anthropic reversed a controversial policy that quietly reduced Claude Fable 5’s capabilities when it detected requests related to frontier AI development. The restriction had been disclosed in the model’s system card, but users were not visibly notified when the intervention occurred. This sparked backlash from researchers, developers, and policy experts who argued that hidden downgrades undermined transparency and trust.

Anthropic said the affected traffic was small, around 0.03%, but later acknowledged that it “made the wrong tradeoff” and said it would make the safeguards visible. The issue highlights a bigger tension in frontier AI: labs want safeguards around high-risk uses, but users expect to know when a model’s performance or behavior is being intentionally limited.

Mistral Bets on Sovereign AI for Regulated European Sectors

Mistral AI is trying to build a sovereign European AI infrastructure business while still distributing its models through Microsoft Azure, Google Cloud, and AWS. This creates a strategic tension: hyperscalers give Mistral reach and revenue today, but may also make it harder to prove the value of independent European AI infrastructure.

The strongest opportunity for Mistral may not be becoming a general-purpose cloud provider. Instead, its advantage could come from serving regulated sectors such as finance, healthcare, government, defense, and public infrastructure that need compliant, auditable, European-jurisdiction AI inference. The key question is whether Mistral can convert its European identity, open-weight model strategy, and regulatory positioning into durable enterprise relationships before US hyperscalers close the sovereignty gap.

EU Cloud and AI Development Act Could Define Sovereign Cloud in Law

The European Commission has proposed the EU Cloud and AI Development Act, or CADA, as part of a broader Tech Sovereignty package. The proposal would turn digital sovereignty into binding procurement rules by introducing an EU-wide sovereignty assessment framework and steering public-sector contracts toward compliant providers. CADA is targeted for adoption in Q4 2027.

The major unresolved issue is how “sovereign cloud” will be defined. A narrow ownership-based definition, supported by CISPE and France, could favor European-owned providers but risk limiting access to high-performance hyperscaler infrastructure. A broader definition could include performance, resilience, and practical capability. The debate is especially important because Europe is already facing compute constraints, while AI workloads are increasing demand for advanced chips and data-center capacity.

Databricks Reportedly Eyes Funding at Up to $175B Valuation

Databricks is reportedly in talks to raise a new funding round that could value the company between $165 billion and $175 billion. This would be a major jump from its earlier $134 billion valuation, following roughly $5 billion in equity financing and around $2 billion in new debt capacity earlier this year.

The company reportedly has a $5.4 billion revenue run rate, up more than 65% year over year, and serves over 20,000 customer organizations. CEO Ali Ghodsi has also reportedly told investors that Databricks still plans to pursue an IPO, potentially as soon as next year. The funding talks show how strongly investors continue to value companies sitting at the center of enterprise data and AI infrastructure.

Europe Sees Over €2.3B in Tech Funding in the First Week of June

European tech saw more than 55 funding deals worth over €2.3 billion in the first week of June, alongside more than 10 exits, M&A transactions, rumors, and related stories. The top funded sectors were analytics, quantum, and travel, while Belgium, the UK, and Spain led by country. Standout deals included Kpler’s $1B+ growth equity investment, Oxford Quantum Circuits’ £260M round, Perk’s €258M credit line, and Factorial’s $150M Series D at a $2.5B valuation.

The week also included notable M&A activity, such as Salesforce acquiring Contentful and Vertice buying Vendr. Overall, the numbers suggest that European tech funding remains active, with strong investor interest in analytics, quantum, AI-native platforms, and enterprise software.