enclaive

Managed Kubernetes is becoming a commodity – how can MSPs stand out? As more companies migrate sensitive workloads to container platforms, there are now numerous Managed Kubernetes services out there. And to the customer, they’re almost interchangeable. At the same time, moving to a Managed Service model still poses security and compliance issues, especially for regulated industries: ❌In traditional architectures, data is not protected during processing ➡️Implicit trust in the provider is still required ➡️Compliance and regulation as dealbreakers But what if your customers no longer had to trust you – because security and confidentiality is technically enforced? This is exactly what our Confidential Computing portfolio enables. In our new Solution Brief, you’ll discover how Managed Confidential Kubernetes helps you: ✅ Protect sensitive customer workloads – even from privileged admins ✅ Clearly differentiate your portfolio ✅ Unlock new high-value, regulated customer segments ✅ Create premium services with higher margins And the best part: ❎ No changes to your preferred current Kubernetes. Why you need to act now: Confidential Computing is rapidly gaining traction, with ~50% CAGR expected over the next 5-10 years. Also, Gartner has named it one of its Top 10 Strategic Technology Trends for 2026 and predicts that by 2029, up to 75% of cloud workloads will be protected “in use”. Any questions? Just contact our Director of Channel Sales, Robert Specht Link to our Managed Kubernetes Solution Brief in the comments 👇

Stay In The Know Brussels just moved the goalposts. On 27 July, the AI Omnibus entered into force. High-risk obligations under the AI Act shift from 2 August 2026 to 2 December 2027. Embedded high-risk systems get until August 2028. The Commission gave its reason plainly: harmonised standards and compliance tools were not ready. Read the delay correctly. The requirements survived intact. Data governance, logging, human oversight, security. You gained 16 months to build evidence, and your auditors gained 16 months to prepare sharper questions. The question that outlasts every deadline shift: where does your AI process sensitive data, and who can read it while the model runs? Confidential computing answers with hardware. Workloads stay encrypted during inference, and the CPU issues a cryptographic attestation that proves it. Teams that build this in now walk into December 2027 with evidence, while others start drafting policies. #AIAct #ConfidentialComputing #AICompliance

The phrase "data sovereignty" appears in almost every cloud contract negotiation, GDPR compliance review, or NIS2 gap assessment. Yet it is rarely defined at an architectural level. Here is the definition that matters for cloud security: Data sovereignty means retaining control over what happens to your data — including when it is being processed in someone else's infrastructure. Not just who can access it legally, but who is technically able. Many organisations assume they have this control because they have a Data Processing Agreement (DPA) – and, where relevant, Standard Contractual Clauses  – with their cloud provider. These agreements create binding obligations, safeguards and legal remedies. They do not, by themselves, remove the provider’s technical ability to access plaintext data. ⁉️ Why this matters: Schrems II exposed the limits of contractual safeguards: contracts cannot bind third-country public authorities or neutralise conflicting access laws. So even with a DPA in place a US cloud provider can still be compelled to disclose data under the CLOUD Act or FISA. Therefore, data sovereignty cannot be achieved through contracts alone.  True data sovereignty means the cloud provider technically cannot access your data. Even if a government compels them. Even if their infrastructure is compromised. 👉🏼 But this requires a different architecture, not a different contract. The architecture has a name: confidential computing. Share with your legal and procurement team. The gap between "DPA" and "sovereignty" is one most organisations haven't formally mapped. #DataSovereignty #GDPR #NIS2 #Schrems2 #ConfidentialComputing #EUCloud

A SOC 2 Type 2 report covers a defined period – often12 months. It is a retrospective: An independent auditor assesses whether specified controls were suitably designed and operated effectively throughout that period... It does not prove those controls are in place today. 🚨Remote attestation is fundamentally different. Attestation is a cryptographically signed statement produced by the hardware itself, in real time, that says: this specific code is running on this specific hardware in this specific configuration. Right now. Not 12 months ago. The statement is protected by a hardware-rooted attestation key and verified through a manufacturer-backed chain of trust .It cannot be produced by software pretending to be trusted hardware. A valid attestation report is a mathematical proof, not a human assessment. ✨The practical implication: A SOC 2 report tells you that a cloud provider’s controls operated effectively during the audit period... It does not tell you whether a privileged administrator accessed your data last Tuesday. An attestation report tells you – before you send a single byte of sensitive data – whether the workload receiving it is the exact code you approved, running on hardware you trust, in a configuration that meets your defined security policy . Don't rely on the auditor's report alone. Verify the workload’s integrity yourself. 👇Share this with your GRC team. Remote attestation is the bedrock for compliance: it turns periodic assessments of security controls into continuously available and verifiable evidence. #RemoteAttestation #ConfidentialComputing #SOC2 #ContinuousCompliance #CISO

How remote attestation works - step by step Attestation is the word everyone in confidential computing uses and almost nobody fully explains. It is the mechanism that answers the question: "How do I know the TEE I'm sending sensitive data to is actually genuine –and hasn't been tampered with?" Here is the actual sequence of operations. #InsideTheEnclave #RemoteAttestation #ZeroTrust

What does "privileged access" actually mean at the hardware level? We talk about "privileged access" as though it is a single concept with a clear boundary. It is not. Privilege in computing is a spectrum – built into the hardware itself – and understanding it reveals why the shared responsibility model has a gap that no policy can close. #InsideTheEnclave #PrivilegedAccess #CloudSecurity

I have asked hundreds of people the same question. Almost nobody can answer it. On 28 June I was a guest at @Everlast AI, talking about 𝗰𝗼𝗻𝗳𝗶𝗱𝗲𝗻𝘁𝗶𝗮𝗹 𝗔𝗜. I have been working on this for over ten years now, using cryptographic techniques like multi-party computation protocols (MPC) or fully homomorphic encryption (FHE). Recently I switched to TEE-based approaches due to their performance and universal applicability. And there is this one moment that happens almost every time. I ask something simple: “Where is your data right now, while the system is actually using it?” And it gets quiet. People know where their data is stored. They know how it moves. But that part in the middle? No idea. And here is the thing: while data is being processed, it is usually just sitting there. Open. Readable. Exposed to whatever runs underneath. Nobody likes to think about that. For normal software, that is already a problem. For AI, it is the whole thing. Because AI only gets useful the moment you feed it your most sensitive data. And that is exactly the moment it gets risky. So the real question was never “Which model do we use?” It is: Who can technically see this data while the system works with it? That is what we talked about. No hype. Just the part most people skip. 𝗧𝗵𝗲 𝗳𝘂𝗹𝗹 𝗰𝗼𝗻𝘃𝗲𝗿𝘀𝗮𝘁𝗶𝗼𝗻 𝗱𝗿𝗼𝗽𝘀 𝘀𝗼𝗼𝗻 𝗼𝗻 @𝗞𝗜𝗕𝘂𝗯𝗯𝗹𝗲. So before you watch, one question: Where is your data least protected right now – and are you sure you actually know? #ConfidentialComputing #CloudSecurity #ZeroTrust #AISecurity #DataSecurity

Very important information for pretty much every financial institution 👇

🥁Behind the scenes: two days in Frankfurt Last week, about 20+ Enclaivians showed up at a hotel outside Frankfurt. Some of us had worked together for months. A few of us had never stood in the same room. enclaive runs remote first. Most of our workday lives in Mattermost threads and email. You learn someone's writing voice long before you learn how they take their coffee. You know who ships fast, who asks the sharp question, who drops the meme at 11pm (you know who you are 😉 ). And yet you have never shaken their hand. So we fixed that.Two days. The agenda said alignment and priorities, and we did that work. We argued about what matters next quarter. We got honest about what slows us down. We left with a shorter list and a clearer why. We all know that the best part happened in the gaps. Over breakfast, when someone turned out to be funnier in person than in any thread. In the hallway, when a problem that had bounced around for a week got solved in four minutes, because we were standing next to each other. At dinner, when titles dropped away and we were just people who build the same thing and enjoy every second of it. One thing stuck with me. You can align a roadmap over video. The trust that carries a team through a hard week comes from sharing a table. Remote first gives us reach, focus, and teammates across borders. It does not hand us belonging. We go make that ourselves. So we did. On Monday, every message landed differently. There was a face behind it now. ✨To every Enclaivian who made the trip, thank you. 🙏🏾 One more note. Thank you Andreas Walbrodt and Sebastian J. Gajek for making this happen. Let's be honest, expertise alone gets you nowhere. It takes leadership and a clear vision to gather the right people in one room and point them the same way. That is the part that turns a group of experts into a team. On that note, let’s keep thriving! #team 💪🏾🥂 #RemoteFirst #TeamBuilding #CompanyCulture #enclaive

Your AI use case is blocked because your data is sensitive. Here's the architectural answer. The use case is clear: use AI to analyse vast amounts of data like patient, financial records, or legal documents. The blocker is equally clear: that data cannot leave your control boundary unencrypted, cannot be visible to the model vendor, cannot be used for training, cannot appear in another customer's output. Most teams respond with policies: "don't paste sensitive data into ChatGPT." This is reasonable guidance, but it’s not an architecture. The real answer is Confidential AI: 1. Local Governance & Pseudonymization As a first layer of defense, highly sensitive entities – names, account numbers, diagnoses, dates of birth – can be tokenized before the prompt leaves your immediate local boundary. "Patient John Smith, DOB 12/03/1978" becomes "Patient [TOKEN_A], DOB [TOKEN_B]". The token mapping lives only inside your boundary. The model receives the structure, ensuring an extra layer of data minimization before processing. 2. Confidential inference runtime The model and your data run inside a Trusted Execution Environment (TEE). While the host infrastructure handles encrypted data (ciphertext), processing happens exclusively inside hardware-isolated memory. The prompt, the context, and the output are never visible to the infrastructure operator in plaintext – even if their infrastructure is compromised or their team is compelled to disclose. 3. Cryptographic Attestation of the runtime Before any data enters the pipeline, the inference runtime is attested. The hardware produces a cryptographically signed attestation report proving that the model is the approved version, running on trusted hardware, in an unmodified state. Only after this proof is verified does your key management service release the decryption keys. Together, these components move security from organizational trust to cryptographic certainty. Your CISO can sign off on a verifiable technical architecture, not just a policy memo. ❓ Which component is blocking your AI deployment? Drop a comment — or tag your Head of AI. #ConfidentialAI #LLMSecurity #GenAI #GDPR #ConfidentialComputing #AIGovernance