
Every few years the ground under computing moves. Right now it is moving in two directions at once, and both run straight through the work security people do every day.

The first shift is the obvious one. Artificial intelligence is climbing into the operator's chair. Models that a year ago could barely string together a working nmap command are now chaining reconnaissance, exploitation, and reporting with uncomfortable competence. So it is fair to ask the question out loud: if a machine is going to do the hacking, what happens to a desktop Linux distribution built so a human could do it by hand?

The second shift is quieter and far larger. The world is pulling apart into separate technology stacks. Whole regions, led by the Asian and Pacific markets, are deliberately stepping away from Western silicon to shrink their exposure to a handful of American monopolies and to pursue true technological sovereignty. The chip carrying that ambition is called RISC-V, and Parrot 7.0, released on December 25, 2025, was the first penetration testing distribution in the world to run on it natively. As of this writing, it remains the only one.

Here is the part worth sitting with before we get anywhere near transistors and toolchains. The chip is the smaller half of the story.

## The part that is easy to miss

If you still picture Parrot as a green-tinted laptop desktop with eight hundred tools stapled on, you are holding the souvenir and missing the engine. Parrot has spent years turning itself inside out. The desktop is still there and still excellent, but the centre of gravity moved a while ago, away from the laptop entirely, toward something closer to a framework, an arsenal that can be picked up and carried to wherever the work is happening.

That work stopped living on laptops some time ago. It lives in the cloud now, in ephemeral infrastructure conjured for the length of an engagement and then evaporated, so the arsenal followed it there. The [Parrot container images](https://www.parrot.run) are not an afterthought conversion of the ISO. They are a first-class, officially sponsored Docker project, OCI images assembled with intent, one layer stacked on the next in the leanest order the format allows, so that a full offensive toolkit or a single tool pulled on its own spins up clean on whatever cloud you happen to be standing on.

Then Parrot connected all of this to the first shift. It built its own MCP harness, [MCPwn](https://mcpwn.parrotsec.ai), a server that provides language models with both low-level primitives and high-level capabilities through a protocol they already understand, running each tool inside an isolated container rather than directly on the host.

This addresses the AI question from earlier. Yes, machines may take over the keyboard. The human in the loop may become less involved, and for some tasks may eventually disappear entirely. But models that perform security work still require tools. In fact, they require them more than ever, and they need those tools in a form they can access directly. This is exactly what Parrot now provides: clean containers that can be deployed and capabilities that can be invoked over MCP. For an AI conducting offensive operations, having an arsenal that can be summoned on demand is critical, and this approach may be the most effective method available.

So the worry inverts. The rise of AI is not what makes a tool distribution irrelevant. It is what makes the tool layer more valuable than the interface that used to sit on top of it.

## Where the two shifts meet

Now put both shifts in the same room, because that room turns out to be RISC-V.

Here is the part almost nobody has connected yet, and it is the one that matters most. RISC-V is not a curiosity or a teaching aid. It is the only credible alternative to Western CPUs in existence, the one architecture a country or a company can adopt to stop renting its silicon from Intel, AMD, or Arm. Sovereign technology stacks need sovereign chips, and the AI workloads those stacks are racing to stand up have to run on those chips too. RISC-V is the only thing on the table that answers both demands at once. So picture the near future plainly. A region builds its own AI cloud on its own processors, then points an AI-driven security platform at a target sitting inside that cloud, and the platform reaches for its tools the way it always does, only to find almost none of them were ever built for the chip running underneath it.

Unless it is running Parrot.

That single fact is the strategic point of this release. A line in a changelog about a new CPU architecture represents a statement about where the field is heading. Parrot now occupies the intersection of two major trends in security work: the move toward AI operators and the move toward open sovereign hardware. It was the first to arrive.

## So what actually is RISC-V

Time to earn the technical half of the title.

At the very bottom of every processor is the instruction set architecture, or ISA: the contract that defines which fundamental operations the chip understands and how software can request them. For decades, this contract has been proprietary. x86 belongs to Intel and AMD. Arm licenses its designs and charges royalties from every implementer. To build a serious CPU, you had to pay for permission and accept that part of your development roadmap remained under someone else's control.

RISC-V changes this. It is an open standard, originally developed as an academic project at UC Berkeley, that anyone can implement without a license or permission. Currently, it is the only processor architecture that is both truly open and seeing global adoption. It was designed to be modular, with a small mandatory base and optional extensions for vector math, cryptography, and the tensor operations that machine learning relies on. This allows designers to include exactly what a workload requires and omit the rest. There are no royalties, no permission required, and no single company controls the architecture.

That last property explains its rapid adoption. RISC-V International, the organization that stewards the standard, deliberately moved its headquarters to Switzerland to remain outside the jurisdiction of any single government. Its membership has grown from a few hundred organizations in 2019 to several thousand today. There are already more than two billion RISC-V cores in the world. SHD Group projects this will exceed 21 billion by 2031, with RISC-V International's 2025 annual report indicating a 33.7% market penetration. Europe is investing public funds in RISC-V supercomputing. In Asia, where the push is explicitly framed as silicon sovereignty, the architecture has moved from research curiosity to national strategy. For many countries, RISC-V is not simply another option but the only realistic path to designing their own processors without external dependencies.

And the silicon is no longer a hobby project. Vendors that began by shipping affordable developer boards have since moved into higher-end markets, where they are now producing RISC-V processors with dedicated neural processing units for running large language models. At FOSDEM 2026 in Brussels, which took place a few months after the Parrot 7.0 release, the Parrot team met with DeepComputing's CEO. DeepComputing unveiled a RISC-V mainboard based on the SpacemiT K3, demonstrating the architecture's progress toward AI workloads. Alongside Tenstorrent, whose RISC-V accelerators deliver substantial performance, the company presented a roadmap covering the full spectrum from edge AI PCs to datacenter-class compute for LLM workloads. DeepComputing has also started providing AI startups with early access to RISC-V workstations and servers. The architecture that was on hobbyist desks two years ago is now being engineered into the clusters that will train and serve the next generation of models.

## The unglamorous problem Parrot quietly solved

Here is where the idealism of open hardware meets the practical challenges of software distribution, and the encounter is not gentle.

A penetration testing distribution is only ever as useful as its arsenal, and much of that arsenal was never designed for RISC-V. Some tools lack any RISC-V support. Many others could theoretically be compiled for it, but theory carries little weight in practice. Porting real software to a new architecture means navigating cross-compilation, build systems that assume x86 or Arm, dependencies that have not been ported, makefiles that must be examined line by line, and compile errors that must be diagnosed repeatedly in a process that may never produce a working binary. Metasploit is a prime example: it requires slow manual rebuilds to run on RISC-V, and other distributions such as the official Metasploit Docker container do not include RISC-V support. However, Metasploit does run on RISC-V when using Parrot.

This is exactly the job that looks, on paper, like a flawless task to hand to an AI, and exactly the job that punishes you for trying. Pointing a language model at a fresh RISC-V box and asking it to bootstrap a toolkit from source means paying, in tokens, for it to parse those build files, interpret a cascade of compiler errors, guess at fixes, then go around the loop again with no guarantee of a working binary at the end. The cost of that loop, run across a real-world toolkit rather than one tidy program, is substantial, and it is precisely the hidden expense that undermines the economics of an automated operation before it has accomplished anything useful.

Parrot's solution is straightforward and practical. The team performed the porting work once and correctly, so others do not need to repeat it under pressure. Released on December 25, 2025, Parrot 7.0 includes a pre-assembled RISC-V root filesystem and provides repository-level support for the architecture. In practical terms, this means the Debian base and all Parrot tools compiled for RISC-V are available and pre-built, accessible through standard package installation. What would cost an AI agent significantly in repeated failed build attempts becomes, with Parrot, a simple one-line installation. There is no need for makefile analysis, compiler error troubleshooting, or hoping that dependencies resolve correctly. The tools are present on the open chip and ready to use.

## The shape of the bet

So let us state this clearly, because this is the crucial point. Artificial intelligence requires tools to perform security work, and Parrot provides exactly those tools. Furthermore, an AI operating on RISC-V requires a software stack compiled for RISC-V. Today, Parrot is the only project in the world that offers this. It was not just one of the first; it remains the only one.

And here is the key detail. Parrot released this solution before the target hardware was even available for purchase. The RISC-V tool support was included in Parrot 7.0 on December 25, 2025, weeks before AI-class RISC-V boards were unveiled and months before they entered the market. The arsenal was prepared and waiting before the machines that require it became accessible.

Others will inevitably follow once the trend becomes undeniable. Parrot recognized it early, made the investment, and delivered a complete solution before demand materialized, rather than reacting after the fact. The desktop interface was never the primary product. The tool arsenal is the product, and it supported RISC-V before the hardware was even widely available.

Parrot Speaks RISC-V