Daily brief at 7am Melbourne. Unsubscribe any time.

Sunday 19 April 2026

A critical RCE flaw in protobuf.js puts millions of JavaScript apps at risk, Grinex blames spies for a $13.7M hack, and Cerebras files for an IPO.

Lead story

Proof-of-Concept Published for Critical RCE Bug in protobuf.js — Patch Now

If your JavaScript stack touches protobuf.js, stop reading this and go check your dependency tree. A critical remote code execution vulnerability in the library has a working proof-of-concept exploit circulating in the wild, which moves this from "patch soon" to "patch today."

What happened. Researchers disclosed a severe flaw in protobuf.js — the JavaScript implementation of Google's Protocol Buffers, a binary serialisation format used everywhere from gRPC microservices to mobile backends. The vulnerability allows an attacker to execute arbitrary JavaScript code, and crucially, someone has already published working exploit code. That's the part that accelerates the clock.

Why this matters. Protocol Buffers are something of an invisible plumbing layer. Developers reach for protobuf.js when they need fast, structured data serialisation — it's embedded in thousands of apps, internal tools, and cloud services without anyone thinking twice about it. The npm package has tens of millions of weekly downloads. That's a massive blast radius for a single library flaw.

The remote code execution angle makes this especially nasty. An attacker who can feed malicious protobuf-encoded data to a vulnerable endpoint can potentially run code in the context of the server-side application — no credentials required, no second step needed. In a microservices architecture, that kind of initial foothold can cascade quickly.

The PoC factor. Security research is one thing; a published proof-of-concept is another. Once working exploit code is public, the barrier to exploitation drops to near zero. Script kiddies, ransomware crews, and nation-state operators can all pick it up. The window between disclosure and active exploitation in the wild has shrunk dramatically over the last few years — sometimes it's measured in hours.

Who's exposed. Any Node.js or browser-based application that uses protobuf.js to deserialise untrusted input is potentially in scope. That includes services accepting protobuf-encoded requests from external clients, internal tools that process data from third-party integrations, and anything in a microservices mesh where one service feeds protobuf data to another.

What to do. Update protobuf.js to the patched version immediately — check the library's GitHub releases and the npm advisory for the exact version number. If you can't patch right now, consider whether you can add input validation or temporarily restrict which clients can submit protobuf-encoded data to affected endpoints. Run a dependency audit across your repos; protobuf.js is often a transitive dependency pulled in by something else, meaning you may not even know you're using it.

What to watch. The next 48–72 hours are critical. Watch for reports of active exploitation in the wild. If you run a bug bounty programme or have a threat intelligence feed, this one is worth flagging internally today. Expect your security tools to start flagging related indicators within the week.

Also today

Sanctioned Crypto Exchange Grinex Blames Western Spies for $13.7M Hack

Grinex, a Kyrgyzstan-based cryptocurrency exchange that was sanctioned by both the UK and US last year, has shut down after losing roughly $13.74 million in what it's calling a state-sponsored cyberattack. The exchange pointed the finger squarely at Western intelligence agencies, though it offered no technical evidence to back that claim. Sanctioned exchanges make a convenient target narrative, and independent verification is essentially impossible here. Still, the shutdown removes a platform that regulators had already flagged as a money-laundering risk — so the outcome may suit everyone regardless of who actually pulled the trigger.

The Hacker News

Cerebras Files for IPO With a $10B OpenAI Deal in Its Back Pocket

AI chip startup Cerebras has filed for an IPO, giving investors a formal look at the books of one of Nvidia's more serious challengers. The timing is notable: Cerebras recently struck a deal with Amazon Web Services to supply chips for AWS data centres, and separately landed an agreement with OpenAI reportedly valued at over $10 billion. For a company that spent years as an interesting-but-niche player, those are serious commercial anchors for a listing. The broader question is whether the AI infrastructure spending wave holds long enough for the IPO to land well.

TechCrunch

Tesla's Robotaxi Expands to Dallas and Houston

Tesla has extended its robotaxi service to Dallas and Houston, making it available across three Texas cities after launching in Austin last year. The company began offering fully driverless rides — no safety operator in the seat — in January 2026. Texas has been a permissive regulatory environment for autonomous vehicles, which explains the concentrated rollout. The bigger signal to watch is whether Tesla can translate this controlled, single-state expansion into a model that holds up at genuine scale, or whether the geography is doing a lot of the heavy lifting.

TechCrunch

Anthropic Is Back in the Room With the Trump Administration

Despite the Pentagon recently flagging Anthropic as a supply-chain risk — an unusual designation for a US-based AI company — the Claude maker is apparently still engaged in conversations with senior Trump administration officials. The thaw matters because US government AI procurement and export control policy are both in flux, and companies locked out of those conversations lose significant influence over how the rules get written. Anthropic's willingness to keep talking despite the Pentagon signal suggests it's prioritising long-term policy access over short-term optics.

TechCrunch

App Store Growth Is Back — and AI Tooling May Be the Reason

New figures from app analytics firm Appfigures show a notable uptick in new app launches on the App Store in 2026, reversing a multi-year slowdown in mobile software releases. The working theory is that AI-assisted development tools are dramatically lowering the cost and time required to ship a functional app, enabling a new wave of solo developers and small teams to publish products that previously would have taken larger resources. If the trend holds, it reframes AI coding tools not just as a productivity gain for enterprise developers, but as a genuine democratisation of mobile software creation.

TechCrunch

Sources consulted