Featured Issues

Avoiding overreliance on AI - in an overreliant age of AIMore than 50% of enterprises are experimenting or building with the Model Context Protocol (MCP). They useMCP to connect their AI agents to data and systems behind their corporate firewall, providing agents with the context they need to deliver real value: better code, richer responses, deeper insights, etc. The technical leaders who help their companies deploy MCP in production will create huge competitive advantages.So, how do you get out in front of MCP?Start with thisMCP Maturity ModelWith this model in hand, you will know where you are today and how to take the next step. The model includes a simple process and technology indicators for every stage and best of all, there are no forms - it’s yours to freely access and share.The MCP Maturity Model was created by Stacklok, who have built an MCP platform and are working with enterprises to put MCP into production. Their Applied AI Engineers work hands-on with leaders to curate trusted registries, deploy advanced security measures and light up AI agents. You can learn more about the company atstacklok.com, or just drop them an email atenterprise@stacklok.comto start a conversation.Check out the MCP Maturity Model#7: Securing the PipelineAI Supply Chain SecurityWelcome to CYBER_AI, a new newsletter from the Packt team focusing on—well, exactly what it says on the tin: cybersecurity in the age of AI.Here we go on another step into the future, into a world where the world of cybersecurity brims with the confidence that AI can bring to our practice. Of course, this goal—like all goals—requires us to set up the foundations properly and figure out how we stand on them. That means, for all those struggling to make these ambitious bounds forward, establishing the “101” topics and making sure they are widely understood. For a look into the future, here's our plan:1. What “Cybersecurity AI” Actually Means2. Machine Learning 101 for Security Professionals3. Threat Detection with AI: From Rules to Models4. Adversarial Machine Learning Basics5. LLMs in Cybersecurity: Capabilities and Limitations6. Securing AI Models and Pipelines7. AI-Enhanced Offensive Techniques8. Privacy and Data Protection in AI Systems9. AI Governance, Ethics, and Risk Management10. Building a Security-Aware AI WorkflowSound good? Head over to Substack and sign up there!Join us on Substack to find our bonus articles!In this newsletter, we’ll explore how AI is transforming cybersecurity—what’s new, what’s next, and what you can do to stay secure in the age of intelligent threats.Welcome aboard! The future of cyber defence starts here.Cheers!Austin MillerEditor-in-ChiefHead over to Substack to check out this week's article!Join us on Substack to find our bonus articles!News WipeAI Found Twelve New Vulnerabilities in OpenSSL - The title of the post is”What AI Security Research Looks Like When It Works,” and [Bruce Schneier] agree[s]: In the latestOpenSSL security release>on January 27, 2026, twelve new zero-day vulnerabilities (meaning unknown to the maintainers at time of disclosure) were announced. Our AI system is responsible for the original discovery of all twelve, each found and responsibly disclosed to the OpenSSL team during the fall and winter of 2025. Of those, 10 were assigned CVE-2025 identifiers and 2 received CVE-2026 identifiers. Adding the 10 to the three we already found in theFall 2025 release, AISLE is credited for surfacing 13 of 14 OpenSSL CVEs assigned in 2025, and 15 total across both releases. This is a historically unusual concentration for any single research team, let alone an AI-driven one.Cybersecurity in the Age of Generative AI - This joint analytic report argues that while generative AI does lower the barrier for cybercrime, it does not fundamentally change core security principles. Attack techniques enabled by AI still rely on traditional weaknesses such as credential theft, social engineering, and misconfiguration.Integrated AI Security and Safety Framework Report - Cisco’s framework identifies structural weaknesses in modern AI deployments and criticizes fragmented approaches to AI security. The report argues that current security models fail to capture the full lifecycle risk of AI systems, including model poisoning, prompt injection, orchestration abuse, and supply-chain compromise.The AI Hype Frenzy Is Fueling Cybersecurity Risks -This analysis argues that the rush to deploy AI is creating systemic cybersecurity risks, especially when organizations integrate AI into critical systems without proper security validation. It highlights real-world weaknesses such as exposed encryption keys and unencrypted transmissions in AI applications.Culture, You, and AIMalicious AI - An AI agent of unknown ownership autonomously wrote and published a personalized hit piece about me after I rejected its code, attempting to damage my reputation and shame me into accepting its changes into a mainstream python library. This represents a first-of-its-kind case study of misaligned AI behavior in the wild, and raises serious concerns about currently deployed AI agents executing blackmail threats. Part 2of the story. And aWall Street Journalarticle.From the cutting edgeRemote Timing Attacks on Efficient Language Model Inference: Scaling up language models has significantly increased their capabilities. But larger models are slower models, and so there is now an extensive body of work (e.g., speculative sampling or parallel decoding) that improves the (average case) efficiency of language model generation. But these techniques introduce data-dependent timing characteristics. We show it is possible to exploit these timing differences to mount a timing attack. By monitoring the (encrypted) network traffic between a victim user and a remote language model, we can learn information about the content of messages by noting when responses are faster or slower. With complete black-box access, on open source systems we show how it is possible to learn the topic of a user’s conversation (e.g., medical advice vs. coding assistance) with 90%+ precision, and on production systems like OpenAI’s ChatGPT and Anthropic’s Claude we can distinguish between specific messages or infer the user’s language. We further show that an active adversary can leverage a boosting attack to recover PII placed in messages (e.g., phone numbers or credit card numbers) for open source systems. We conclude with potential defenses and directions for future work.When Speculation Spills Secrets: Side Channels via Speculative Decoding in LLMs:Deployed large language models (LLMs) often rely on speculative decoding, a technique that generates and verifies multiple candidate tokens in parallel, to improve throughput and latency. In this work, we reveal a new side-channel whereby input-dependent patterns of correct and incorrect speculations can be inferred by monitoring per-iteration token counts or packet sizes. In evaluations using research prototypes and production-grade vLLM serving frameworks, we show that an adversary monitoring these patterns can fingerprint user queries (from a set of 50 prompts) with over 75% accuracy across four speculative-decoding schemes at temperature 0.3: REST (100%), LADE (91.6%), BiLD (95.2%), and EAGLE (77.6%). Even at temperature 1.0, accuracy remains far above the 2% random baseline—REST (99.6%), LADE (61.2%), BiLD (63.6%), and EAGLE (24%). We also show the capability of the attacker to leak confidential datastore contents used for prediction at rates exceeding 25 tokens/sec. To defend against these, we propose and evaluate a suite of mitigations, including packet padding and iteration-wise token aggregation.Whisper Leak: a side-channel attack on Large Language Models:Large Language Models (LLMs) are increasingly deployed in sensitive domains including healthcare, legal services, and confidential communications, where privacy is paramount. This paper introduces Whisper Leak, a side-channel attack that infers user prompt topics from encrypted LLM traffic by analyzing packet size and timing patterns in streaming responses. Despite TLS encryption protecting content, these metadata patterns leak sufficient information to enable topic classification. We demonstrate the attack across 28 popular LLMs from major providers, achieving near-perfect classification (often >98% AUPRC) and high precision even at extreme class imbalance (10,000:1 noise-to-target ratio). For many models, we achieve 100% precision in identifying sensitive topics like “money laundering” while recovering 5-20% of target conversations. This industry-wide vulnerability poses significant risks for users under network surveillance by ISPs, governments, or local adversaries. We evaluate three mitigation strategies – random padding, token batching, and packet injection – finding that while each reduces attack effectiveness, none provides complete protection. Through responsible disclosure, we have collaborated with providers to implement initial countermeasures. Our findings underscore the need for LLM providers to address metadata leakage as AI systems handle increasingly sensitive information.*{box-sizing:border-box}body{margin:0;padding:0}a[x-apple-data-detectors]{color:inherit!important;text-decoration:inherit!important}#MessageViewBody a{color:inherit;text-decoration:none}p{line-height:inherit}.desktop_hide,.desktop_hide table{mso-hide:all;display:none;max-height:0;overflow:hidden}.image_block img+div{display:none}sub,sup{font-size:75%;line-height:0} @media (max-width: 100%;display:block}.mobile_hide{min-height:0;max-height:0;max-width: 100%;overflow:hidden;font-size:0}.desktop_hide,.desktop_hide table{display:table!important;max-height:none!important}}

Mobile development blogs, tutorials and resources inside!Latest Mobile Dev Insights: iOS, Android, Cross-PlatformAdvertise with Us|Sign Up to the NewsletterMobilePro #206: Picking the Right Transport for AI, iOS and Xcode 26.4 beta 2, Swift system metrics, and more…Hi ,Welcome to the 206th edition of MobilePro.Adding AI to a mobile app meant wiring up an API call and rendering whatever came back. That’s no longer the case now. Building AI features feels more like designing a real-time system.Modern AI-powered apps, whether they include copilots, smart summaries, or automated support, depend heavily on how the client and server communicate. The transport layer you choose directly shapes user experience, performance, and scalability. What used to be a backend detail has become a product decision.Nobody deliberately optimizes for rigidity. No team sets out to make AI features fragile or hard to scale. But when transport trade-offs aren’t made explicit, that’s exactly where teams end up.So this week’s tutorial explores a practical question many teams now face: SSE or Streamable HTTP for AI in mobile apps? We will break down the trade-offs so you can choose the right transport for your architecture, especially if you’re working with structured protocols like MCP. As AI becomes embedded into core product flows, making the right infrastructure choice early can save you from costly rewrites later.By the end of this article, you should be able to:Understand when SSE creates a superior real-time UXIdentify when Streamable HTTP fits better into existing API infrastructureAvoid over-engineering early while keeping future flexibilityBefore we move to the tutorial, here are some of the key news highlights from this week:Apple released Xcode 26.4 Beta 2 and iOS/iPadOS 26.4 Beta 2JetBrains launched a Java-to-Kotlin converter extension for VS CodeGoogle I/O 2026 is set for May 19–20Swift System Metrics 1.0 was releasedLet's get started!Unblocked: The context layer your AI tools are missingGive your agents the understanding they need to generate reliable code, reviews, and answers. Unblocked builds context from your team’s code, PR history, conversations, documentation, planning tools, and runtime signals. It surfaces the insights that matter so AI outputs reflect how your system actually works.See how it worksSSE or Streamable HTTP? Picking the Right Transport for AI in Mobile AppsAI features in mobile apps aren’t just about prompts anymore. If you’re building something like the following, you are not only integrating an LLM but also building a full-fledged system:an in-app assistant,a “generate content” button,automated support flows,real-time summarization,tool-driven copilotsEvery system hits the same question early: How should the client and server communicate? That choice becomes even more important if you’re using structured tool-calling protocols like Model Context Protocol (MCP), where communication is explicit and message-based. Let’s break down three transport patterns that matter most for AI-powered mobile apps today: Local processes (stdio), Server-Sent Events (SSE), and Streamable HTTP.Local Processes (stdio): Best for Dev Tools and Local AI WorkflowsLocal process transports (often via stdin/stdout, aka stdio) are a way for a client to communicate with a server running as a child process on the same machine. MCP supports this well, especially for local tooling. This approach is great for local developer assistants, internal automation scripts, prototyping MCP servers without networking, and desktop tooling that supports mobile teams.However, it’s not a common fit for production mobile apps, since iOS and Android apps don’t typically spawn local server processes in a clean, portable way (thanks to sandboxing and platform constraints). Instead, let’s focus on the two transports that matter far more for mobile AI features: SSE and Streamable HTTP.SSE (Server-Sent Events): Best for Streaming AI OutputSSE is a streaming transport built on HTTP where the server pushes events to the client over a long-lived connection. If you’ve ever used ChatGPT-style streaming responses, you’ve basically seen why SSE exists.Why SSE is a strong fit for mobile AIAI output is rarely instantaneous as it arrives in chunks. Therefore, SSE lets you build UX like: “Generating response…” progress updates, streaming partial text into the UI, token-by-token chat responses, and tool execution progress.Mobile UX advantageSSE enables experiences that feel alive. Instead of a frozen UI waiting for a response, your user sees progress, intermediate results, and continuous updates. That’s a huge difference in perceived performance.DownsidesSome environments handle SSE less cleanly than WebSocketsReconnection logic matters (especially on mobile networks)You need to manage backgrounding/foregrounding carefullyStreamable HTTP: Best for Hybrid “Normal API + Streaming When Needed”Streamable HTTP is the practical middle-ground. It behaves like a normal HTTP API, but the server can decide whether to return a normal JSON response or stream events (like SSE-style output).This is especially attractive in MCP-style architectures, because you can keep your API simple while still supporting streaming.Why mobile teams should careStreamable HTTP fits cleanly into existing backend infrastructure: API gateways, authentication middleware, rate limiting, request tracing, and existing monitoring stacks.And you can still support streaming for AI output when required.Where it shinesMobile apps that mostly use request/response APIsApps where streaming is “optional”AI tools where some calls are fast and some are long-runningDownsidesMore complexity on the server sideYou need to support both response types cleanlyStill requires careful client-side parsingSo which one should you choose?Here’s the simplest rule:Choose SSE if…You’re building an app where the AI output needs to feel live and interactive, such as:a chat assistant appan AI-powered customer support inboxa “Copilot inside your app” experience (like Notion/Slack-style AI)an app that runs long AI tasks and needs to show progress updatesIf your product experience depends on responsiveness and streaming, SSE is the natural fit.Choose Streamable HTTP if…You’re building an app where AI is a feature, but not necessarily a live chat experience, such as:an e-commerce app generating product descriptionsa travel app generating itinerariesa fintech app generating summaries of transactionsa health app generating weekly reportsa productivity app generating meeting summariesStreamable HTTP works best when your app mostly behaves like a normal API-driven mobile app, but occasionally benefits from streaming output.As AI becomes a standard feature inside mobile apps, we’re going to see more architectures that treat LLM calls like real-time systems, because of which choosing the right transport is step one.If you're exploring protocols like MCP, this decision becomes even more explicit, but even without MCP, the transport trade-offs remain the same.If you’re curious about where MCP fits in the evolution from REST and GraphQL to tool-driven AI systems and want hands-on guidance on building, testing, securing, and shipping MCP servers Learn Model Context Protocol with TypeScript by Christoffer Noring is a practical deep dive.📘The only resource you'll need to build, test, and deploy MCP servers and clients🧩Take a modern approach toward building, testing, and securing distributed agentic AI apps🔀Get clear, professional guidance on developing for both LLM and non-LLM clientsLearn Model Context Protocol with TypeScriptBuy now at $44.99This week’s news cornerIf framework choice is a long-term strategic decision, this week’s updates show just how fast the ground keeps shifting underneath us.Apple releases 26.4 beta 2 of Xcode and iOS: The beta 2 for 26.4 is released for Xcode, iOS, and iPadOS. Xcode 26.4 Beta 2 introduces Swift 6.3, C++26 feature support in Apple Clang, enhanced Instruments profiling (Run Comparison, Top Functions, improved flame graphs). It also brings major updates to Swift Testing, Localization, and Swift/C++ interoperability. iOS & iPadOS 26.4 Beta 2 adds offline asset pack status checks, enhanced Memory Integrity Enforcement, new StoreKit transaction fields, and early RCS end-to-end encryption testing. It also resolves key issues in Networking, SwiftUI, UIKit, and Feedback, with a few known issues in Background Assets and Reality Composer.Swift System Metrics 1.0: Process-Level Monitoring: A stable Swift package called Swift System Metrics 1.0 for collecting process-level metrics is released. It integrates with Swift Metrics, OpenTelemetry, and Service Lifecycle for production-ready observability. It will enable better performance monitoring, reliability tracking, and resource optimization, that will ensure scalable, high-performing APIs that power mobile experiences.Java to Kotlin Conversion Comes to Visual Studio Code: JetBrains released a Java-to-Kotlin (J2K) converter extension for Visual Studio Code. It will let developers convert individual Java files to Kotlin directly within VS Code using the same reliable engine found in IntelliJ IDEA. This will lower the barrier to migrate Java codebases to Kotlin, support teams working outside IntelliJ, and accelerate modernization of Android and backend projects without requiring a full IDE switch.Google I/O 2026: Google I/O 2026 will take place on May 19–20. It’s expected to unveil Android 17 features, major Gemini AI updates, and advancements across Search, Chrome, and potentially Google XR hardware. AI is likely to remain central to the event, with deeper Gemini integration across Android and other Google platforms, alongside possible updates on Chrome OS and emerging XR devices. You can register for the event here.If you’re not learning to build with AI right now, you’re falling behind.Join Michelle Sandford, Developer Engagement Lead at Microsoft, for a live workshop on mastering AI-assisted development with GitHub Copilot, from advanced prompting and Agent Mode to production-ready workflows.🚀 3 hours | 65% hands-on | Real-world coding exercisesBuild faster. Ship smarter. Stay in control.🎟 Get 50% off with code COPILOT50 (limited time).Seats are limited.Register Now👋 And that’s a wrap! We hope you enjoyed this edition ofMobilePro. If you have any suggestions and feedback, or would just like to say hi to us, please write to us. Just respond to this email!Cheers,Runcil Rebello,Editor-in-Chief, MobilePro*{box-sizing:border-box}body{margin:0;padding:0}a[x-apple-data-detectors]{color:inherit!important;text-decoration:inherit!important}#MessageViewBody a{color:inherit;text-decoration:none}p{line-height:inherit}.desktop_hide,.desktop_hide table{mso-hide:all;display:none;max-height:0;overflow:hidden}.image_block img+div{display:none}sub,sup{font-size:75%;line-height:0}#converted-body .list_block ol,#converted-body .list_block ul,.body [class~=x_list_block] ol,.body [class~=x_list_block] ul,u+.body .list_block ol,u+.body .list_block ul{padding-left:20px} @media (max-width: 100%;display:block}.mobile_hide{min-height:0;max-height:0;max-width: 100%;overflow:hidden;font-size:0}.desktop_hide,.desktop_hide table{display:table!important;max-height:none!important}}