Google and Microsoft have long ruled the workplace productivity game, but, at the same time, they’ve built empires on your data, and Proton is done watching from the sidelines. On March 31, 2026, Proton officially launched its own workspace alternative, called Proton Workspace (they couldn’t have been more direct).

The Proton Workspace, as the company states, is a fully bundled productivity suite designed to give businesses a genuine, encryption-first alternative to those from the big tech companies. 

What’s so unique about Proton Workspace?

Proton’s Workspace includes several useful tools that people already use as part of their digital lives. They include Proton Mail, Calendar, Drive, Docs, and Sheets, VPN, and Pass. It comes in two tiers: Standard at $12.99 per month (billed annually) and Premium at $19.99 per month.

With the latter, you’ll also get 3TB of storage, higher video calling limits, and access to Lumo, the company’s privacy-focused AI assistant. The standout feature, for me, is Proton Meet, a brand-new encrypted video conferencing tool. 

It uses end-to-end encryption via the Messaging Layer Security (MLS) protocol. In other words, not even Proton can listen in. Even if you don’t have an account, Proton says that anyone can host or join a meeting. This, in my opinion, makes the tool quite useful for journalists, activists, or anyone who needs to keep their conversations off the grid. 

It’s no secret that 2025 was the golden age of budget phones. A year where you could get fast performance and capable cameras all for under ₹20k. Unfortunately, 2026 isn’t shaping up to be the year we all hoped it would be. It’s the year we’re starting to feel the consequences of AI in the form of sky-high RAM and storage prices. As you may have guessed, budget phones have fared the worst, given their already thin margins. OPPO has historically done well with its K-series, and this year the Chinese smartphone maker is back with the all-new K14 5G.

The K14 is an interesting phone, to say the least. It packs the Dimensity 6300 processor, a 50MP main camera, and a new design. The first point of comparison for many would be its predecessor, the K13. It was the favourite of many and still comes out as a better phone than the K14. But to me, that’s not the correct way to judge. Sure, the K13 launched at the same price of ₹17,999, but it now costs over ₹20k, so the two devices simply cannot be compared. This had me wondering: How does the new OPPO K14 justify its price tag? To answer this question, I asked OPPO to send the K14. This review sums up my experience with the device over the last two weeks, highlighting its pros and cons to help you make an informed decision.

Design & Hardware

Design has always been OPPO’s bread and butter, with some handsome-looking phones throughout its history. I can pretty much say the same things about the K14 as well. It’s a tried-and-tested design that just works for everyone. The white variant I received has geometric patterns on the back that catch the light in different ways. This adds a bit of style without being overly flashy. I handed the phone to some of my family and friends, and they all had the same opinion. The matte finish picks up no fingerprints, so the phone stays pristine at all times.

The sides are plastic, of course, but that’s not a problem. The silver finish matches the white frame well, and the in-hand feel is good, too. Don’t get me wrong, the K14 is a big phone, but with a slight curve to the frame, it’s manageable enough to use with one hand for someone with average-sized fingers. The back also includes the dual camera setup tucked inside an island. The design is fairly straightforward, but what I do love is that the camera lenses don’t protrude much, so the K14 doesn’t wobble on a table. The weight can range from 209–216g, depending on your color, and it’s distributed pretty well for a phone with a 7,000 mAh battery pack.

Security is handled by a side-mounted fingerprint scanner. It’s plenty fast and works with both wet and greasy fingers, unlike the optical ones. The K14 also gets the myriad of IP ratings, including IP68 and IP69, meaning you can go swimming with the K14 if desired. Though I would always be against that since water damage isn’t covered under warranty. Also, OPPO is heavily advertising the K14’s durability, which I put to the test when I accidentally dropped it from waist height. There was no damage whatsoever.

Display & Speakers

The OPPO K14 features a 6.75-inch HD+ LCD display with a 120Hz refresh rate. I’ll say it straight: the display isn’t a selling point for the K14. The 120Hz refresh rate is great, but the HD+ resolution simply limits everyday usability, and the text does feel noticeably less sharp. However, if you haven’t been testing flagship phones every month, then the K14 would feel right at home. The 1,125-nit peak brightness makes the screen visible outdoors without me having to squint much. The content experience is also fine, but nothing special.

The stereo speakers, on the other hand, are really good. There’s ample separation between instruments, and mids stay decently clear along the spectrum.

Performance

Performance is what makes or breaks the budget smartphone experience. The K14 features the Dimensity 6300 processor, featuring 2 Cortex-A76 cores running at 2400 MHz and 6 Cortex-A55 cores at 2000 MHz, along with the Mali-G57 MP2. In addition, my review unit came with 8GB of LPDDR4x RAM and 256GB of UFS 2.1 internal storage.

The Dimensity 6300 has been around for a long time and has proven to be plenty capable. The K14’s everyday performance is actually really good. ColorOS is my favourite Android skin, and that pays enough dividends. Both light apps and heavier ones like the camera open quickly, switching between tasks is smooth, and the RAM is enough to keep multiple tasks running at the same time.

The animations, while not as pretty on my Find X8, are still fluid enough to make the experience feel premium. The phone also gets a dose of my favourite AI features, including the likes of AI Portrait Retouching, Recompose, and Perfect Shot. All of them help you tweak the camera’s output for a better-looking photo. All that being said, there are many pre-installed apps, including the annoying “Hot Apps” folder, which you’ll need to remove manually.

Since this is a review, I also put the Dimensity 6300 through its paces with a series of benchmarks. The K14 scored 814 in Geekbench’s single-core and 1,981 in the multi-core test. In AnTuTu, the phone achieved 688,325 points. Moving on to gaming, I installed BGMI and played it at Smooth and Extreme settings. Overall, the experience was decent for the price, with stable 60 FPS gameplay and no dropped frames, even in intense situations.

Battery Life & Charging

If you ask me for just one reason to consider the K14, I’ll probably say it’s the 7,000 mAh battery. I couldn’t push the phone hard enough to drain it in a day, and that’s big praise. For context, I started my day at around 9 AM with 100% charge remaining. After my morning cup of coffee, I decided to run benchmarks on the K14. Once that was in order, I watched a few YouTube videos and hopped into a long BGMI gaming session with my friends. The evening was the usual struggle at the gym with the OPPO Enco Air 3 Pro+ connected to the phone. Lastly, I went shopping with my friends and took a few sample photos.

I ended that day with roughly 40% charge remaining. When I did eventually run out of juice the next day, the 45W fast charging came to the rescue. It can charge the battery from 0% to full in around 90 minutes. You can also use the K14 as a power bank to charge other phones with a cable.

Cameras

The K14’s cameras are headlined by a 50MP OV50D sensor paired with a 2MP depth sensor. Selfies are handled by an 8MP shooter. I didn’t expect much from the K14’s cameras, but they are better than expected. Give the main sensor enough light, and it captures some pretty photos, with ample sharpness and decent HDR. The same was true of daytime portraits, where edge detection was on point, and the shallow depth of field looked natural. Skin tones are also managed well. There were a few situations where highlights were slightly blown out, but that isn’t a big issue.

Nighttime is where the OmniVision sensor struggles. Don’t get me wrong, the images are decent, but they just lack details, and there’s noticeable noise. You can also expect to wait before the camera captures the image, since there’s a shutter delay at night, which can make it difficult to capture moving subjects. That said, the 8MP selfie shooter is more than enough to preserve detail while keeping skin tones natural. Videos being capped at just 1080p at 60FPS is a bummer, though.

Verdict

At ₹17,999, the OPPO K14 is the new reality of budget phones in India, and it’s something buyers will have to adjust to. The K14 is a good phone with a pretty design that’s more than capable of handling everyday tasks and even gaming, for that matter. However, it’s not perfect. I’d love to see better nighttime camera performance and a higher-resolution display, but neither is a deal-breaker.

Tired of the stale, fetid air looming over your apartment like a cloud? Check out the Coway Airmega Mighty, an already wallet-friendly home air purifier that’s even cheaper right now as part of the Amazon Big Spring Sale. It’s currently marked down to just $154, a $76 discount from its typical price, but you’ll want to move quickly if you’re interested, as the deal is only available for a limited time.

Coway

Airmega Mighty AP-1512HH

Despite its low price tag and squat stature, the Airmega Mighty is capable of cleaning a substantial amount of space. At full bore, it can handle a 361-square-foot space, although you’ll get the best performance, and save your ears, if you’re closer to a 200-square-foot room. If you don’t want it running constantly, there are built-in timers to automatically shut off after 1, 4, or 8 hours, or you can use Eco Mode, which will run until the Might doesn’t sense any dirty air for half an hour.

That’s right, the Airmega Mighty has a built-in air quality sensor, and it reflects the current state of the air quality using a colored light with three levels. It uses those readings to automatically adjust the fan speed and timing settings on the fly, as well as giving you a peak into how bad the air you’re breathing right now is for you. While it lacks integration with smart home setups like Google Home, it makes up for it by handling all of its own business without Wi-Fi or extra apps on your phone.

While the Coway Airmega Mighty is available in three colors, only the black and silver model is currently discounted, so you’ll have to pay full price if it doesn’t match your living room’s color scheme. We’ve put in the work testing every air purifier we could get our hands on, so make sure to check out the full guide if you’re trying to clean up your space. The Coway is discounted as part of Amazon’s Big Spring Sale, and we’ve got the best deals from products we’ve tested gathered in one place if you want to save some bucks.

“Good phones are getting cheap, and cheap phones are getting good” is the saying we’ve all heard from our good friend MKBHD. It’s a statement that’s stood the test of time, simply because budget phones have improved in performance, camera quality, and overall usability every year. Unfortunately, 2026 feels like a bit of an outlier. It’s the year of AI consequences, with rising component costs that have wreaked havoc in the smartphone world. As you may have already guessed, budget phones are the most affected. They already have razor-thin margins, and these new costs have driven prices up significantly. vivo’s T-series was one of my favourite budget phones of last year, because they understood the formula well. This year, vivo is back again with the T5x.

So, what is the new vivo T5x about? It’s about a new processor, the Dimensity 7400, a full HD+ display with a 120Hz refresh rate, and the ability to record 4K videos. All this at a price of ₹18,999 made me wonder where vivo has cut corners? To find out, I got the new T5x, swapped my SIM to it, and tested it for a couple of weeks. Here’s my experience.

Design & Hardware

The previous T4x was a handsome phone, with a big camera module and a pretty design. And I can pretty much say the same things about the T5x, except for the fact that its design is more mature. vivo’s T-series phones were some of my favourite budget phones. The back captures the light at different angles, making a wavy pattern. This matte finish picks up zero fingerprints, so you don’t have to spend hours cleaning the phone.

In 2026, everyone’s making phones with flat sides. That’s the trend, but people like me do miss the way backs melted into the frames, which was super comfortable. Well, vivo has apparently listened to us, since the back of the T5x has a subtle curve that blends into the frame. It’s a small touch that pays dividends in the comfort department. I had no issues using the phone for long hours or holding it while taking my evening walks. The back also houses the dual-camera setup, and its design resembles the V70’s, with a square camera island, and I quite like it.

The frame is plastic, of course, but that’s not necessarily a bad thing. They are flat, which is good, but the corners are rounded, so they do not dig into your palms. The 215g weight is surprisingly low for a 7,000 mAh battery, and the center of gravity is perfectly centered.

Beyond the basics, the vivo T5x features a side-mounted fingerprint scanner that works well with both dirty and wet fingers. Another major selling point of the T5x is the IP ratings. The device supports IP68, IP69, and IP69+. All these fancy terms mean that you can submerge your phone underwater for over 30 minutes. As always, I’d warn against doing so, since water damage isn’t covered under warranty and nobody wants a dead phone.

Display & Speakers

The vivo T5x features a 6.76-inch 2344×1080 FHD+ display with a 120Hz refresh rate. Yes, the panel is LCD, meaning you won’t get the inky deep blacks, but it fared pretty well in my testing. I used the phone both indoors and outdoors, where the 1200-nit peak brightness kept everything legible without me needing to squint. The panel covers about 83% of the NTSC color gamut, which means plenty of accurate colors for watching Netflix or YouTube. Speaking of the speakers, they are decent. I wouldn’t describe the audio quality as full, but it’s okay if you just want to listen to something on the go.

What I’m not a big fan of, though, are the bezels. They are uneven, and the chin at the bottom is fairly large for the price. In other news, vivo hasn’t stated the type of glass used for protection. Fortunately, there’s a pre-installed screen protector, so I’d recommend keeping that on.

Performance

Performance is what makes or breaks the experience on many budget phones. Fortunately, vivo hasn’t fumbled this with the T5x. Under the hood, the T5x houses the MediaTek Dimensity 7400-Turbo processor, an octa-core chip built on the 4-nanometer process technology. Beyond that, my review unit came with 8GB of LPDDR4X RAM and 256GB of UFS 3.1 storage.

I’ve enjoyed using the new OriginOS skin on several vivo phones this year. Interestingly, none of them have been budget smartphones, so I was quite excited to see how vivo has optimized its UI. The answer? Pretty well. The T5x, coupled with the 120Hz refresh rate, breezes through opening and closing multiple apps without a hitch. The animations, while not as smooth as on the X300 Pro, are still fluid enough. The 8GB of RAM is also enough to hold multiple apps in memory. There’s also the AI eraser that helps effectively remove people from the shot.

That being said, there are some pre-installed apps and the infamous “Hot Apps” folder, though everything can be removed in a matter of minutes. vivo has promised two years of major Android updates and four years of security patches.

To push the Dimensity 7400-Turbo to the limits, I also ran a series of benchmarks. The T5x scored 1,039 on Geekbench’s single-core test and 2,833 on the multi-core test. On the flip side, in AnTuTu, the device scored 1,189,871 points. But benchmarks rarely tell the full story, so I also ran a series of games, including BGMI. In Smooth+Extreme settings, the phone maintained a steady 60 fps gameplay without dropping frames. The phone also includes gaming-centric features such as 4D vibration (exclusive to BGMI), bypass charging, and the ability to cycle through different performance modes.

Battery & Charging

If you need one reason to buy the vivo T5x, it’s probably the mammoth 7,200 mAh battery. It’s a phone that’s impossible to kill in a day. Trust me, I’ve tried. On the first day after I charged the phone to full, I went shopping with my family, where I took about 15-20 photos and left the camera running for more than 30 minutes. After I got home, I played a couple of BGMI games with my friends for an hour, then did a gym session with my Bluetooth earbuds connected to my phone. I ended the day with my favourite activity — doomscrolling on Instagram — and still had about 45% of the charge remaining.

When I ran out of juice, the 44W fast charging came to the rescue. It helped recharge the battery from 20% to 80% in just under an hour, which is plenty fast.

Cameras

Given the current situation, I honestly didn’t expect much from the T5x’s cameras, since corners have to be cut to accommodate the high RAM prices. But the T5x’s cameras are more than good enough for the price. The camera setup includes a 50MP Sony IMX852 main sensor paired with a 2MP depth sensor. Selfies are handled by a 32MP shooter.

As evident from the samples, the sensor produces pleasing images with ample details, accurate colors, and good-enough HDR in daylight. vivo’s portrait mastery also shines on the T5x, with photos that look aesthetic and fairly accurate edge detection in both daytime and nighttime conditions. There were a few situations where highlights were slightly blown out, but that isn’t a big issue.

Speaking of nighttime, I did notice some noise creeping in when capturing the night sky. There’s also a shutter delay at night, so capturing moving subjects could be an issue. That said, after giving the sensor enough time, the results were detailed enough, and the colors weren’t washed out. Selfies, on the other hand, were solid in both daytime and nighttime scenarios, and I had no complaints. Beyond that, I love the 4K video support on both the front and back cameras. The video quality isn’t the best when compared to more expensive phones, but it’s decent enough to capture a family gathering.

Verdict

At ₹18,999, the vivo T5x surprisingly gets a lot of things right. The design looks great, especially in the silver finish. The display, while not OLED, is serviceable and bright in all conditions, and the performance is more than good enough. Not to mention the stellar battery life that lasts more than a day, and cameras that actually capture good photos. Sure, it’s not all perfect, but given the current market conditions, it’s certainly one of the better budget phones I’ve tested this year. And if you’re shopping in the segment, you should check it out.

A series of letters sent by autonomous-vehicle (AV) developers to Democratic US senator Ed Markey of Massachusetts sheds the most light yet on the human side of robot vehicle operations. In the documents, submitted to Markey as part of an investigation into self-driving-vehicle technology and released on Tuesday, seven companies, including Tesla, Amazon-owned Zoox, and Uber- and Nvidia-funded Nuro, released new details about their “remote assistance” programs.

All the companies that responded to the senator’s office say they use remote assistants—humans charged with responding to autonomous vehicles when they get confused, stuck, or in emergencies. The programs, experts say, are an important part of any autonomous vehicle company’s safety considerations, a backstop for a technology that’s becoming safer by the year but will continue to run into new situations on the road indefinitely.

In a report also released Tuesday, Senator Markey said the new details were not enough. “Every autonomous-vehicle company refused to disclose how often their AVs require assistance from [remote assistants]—hiding key information from the public about their AV’s true level of autonomy,” he wrote. “This information is critical for lawmakers, regulators, and the public to understand the potential safety risks with AVs.”

Markey called on the nation’s top federal road safety regulator to look more closely into autonomous vehicle companies’ remote assistance programs, and said he would soon introduce legislation responding to the “safety gaps” his investigation found.

Remote-Controlled Robotaxis

The responses from the autonomous vehicle developers show that, in one critical way, Tesla is an industry outlier. Six of the firms insisted that their remote assistance workers, who work across the US and even, in the case of Waymo, in the Philippines, never actually drive the vehicles directly. Instead, the humans provide input that the autonomous vehicle software then decides to use or ignore.

Not so for Tesla. “As a redundancy measure in rare cases … [remote assistance operators] are authorized to temporarily assume direct vehicle control as the final escalation maneuver after all other available intervention actions have been exhausted,” Karen Steakley, Tesla’s director of public policy and business development, wrote to the senator. The automaker’s remote assistance workers can “take temporary control of the vehicle” at speeds up to or less than 2 mph and can remotely drive a Tesla Robotaxi at up to 10 mph if the vehicle’s software permits it to do so, Steakley said. “This capability enables Tesla to promptly move a vehicle that may be in a compromising position,” she wrote.

Tesla, which has pivoted its business away from making cars and toward autonomous vehicle technology and robots, launched a small ride-hailing service in Austin, Texas, last June. In most of the 50 or so so-called robotaxis operating today, human safety operators sit in the front passenger seats, ready to take over or intervene if something goes wrong. A handful of the vehicles reportedly operate without safety operators. The automaker says its remote assistants are based in Austin and Palo Alto, California.

Autonomous vehicle developers usually avoid direct remote control of their vehicles for several reasons. Small delays between what a human remote assistant is seeing and what’s happening on the road in real time, even by just a few hundred milliseconds, can lead to slower reaction times, an issue exacerbated by network latency. This increases the potential for accidents. “Your ability to drive a car without being in the car is only as stable as the internet connection that connects you to it,” a self-driving-vehicle engineer told WIRED last year.

Anthropic appears to have accidentally revealed the inner workings of one of its most popular and lucrative AI products, the agentic AI harness Claude Code, to the public.

A 59.8 MB JavaScript source map file (.map), intended for internal debugging, was inadvertently included in version 2.1.88 of the @anthropic-ai/claude-code package on the public npm registry pushed live earlier this morning.

By 4:23 am ET, Chaofan Shou (@Fried_rice), an intern at Solayer Labs, broadcasted the discovery on X (formerly Twitter). The post, which included a direct download link to a hosted archive, acted as a digital flare. Within hours, the ~512,000-line TypeScript codebase was mirrored across GitHub and analyzed by thousands of developers.

For Anthropic, a company currently riding a meteoric rise with a reported $19 billion annualized revenue run-rate as of March 2026, the leak is more than a security lapse; it is a strategic hemorrhage of intellectual property.The timing is particularly critical given the commercial velocity of the product.

Market data indicates that Claude Code alone has achieved an annualized recurring revenue (ARR) of $2.5 billion, a figure that has more than doubled since the beginning of the year.

With enterprise adoption accounting for 80% of its revenue, the leak provides competitors—from established giants to nimble rivals like Cursor—a literal blueprint for how to build a high-agency, reliable, and commercially viable AI agent.

We’ve reached out to Anthropic for an official statement on the leak and will update when we hear back.

The anatomy of agentic memory

The most significant takeaway for competitors lies in how Anthropic solved “context entropy”—the tendency for AI agents to become confused or hallucinatory as long-running sessions grow in complexity.

The leaked source reveals a sophisticated, three-layer memory architecture that moves away from traditional “store-everything” retrieval.

As analyzed by developers like @himanshustwts, the architecture utilizes a “Self-Healing Memory” system.

At its core is MEMORY.md, a lightweight index of pointers (~150 characters per line) that is perpetually loaded into the context. This index does not store data; it stores locations.

Actual project knowledge is distributed across “topic files” fetched on-demand, while raw transcripts are never fully read back into the context, but merely “grep’d” for specific identifiers.

This “Strict Write Discipline”—where the agent must update its index only after a successful file write—prevents the model from polluting its context with failed attempts.

For competitors, the “blueprint” is clear: build a skeptical memory. The code confirms that Anthropic’s agents are instructed to treat their own memory as a “hint,” requiring the model to verify facts against the actual codebase before proceeding.

KAIROS and the autonomous daemon

The leak also pulls back the curtain on “KAIROS,” the Ancient Greek concept of “at the right time,” a feature flag mentioned over 150 times in the source. KAIROS represents a fundamental shift in user experience: an autonomous daemon mode.

While current AI tools are largely reactive, KAIROS allows Claude Code to operate as an always-on background agent. It handles background sessions and employs a process called autoDream.

In this mode, the agent performs “memory consolidation” while the user is idle. The autoDream logic merges disparate observations, removes logical contradictions, and converts vague insights into absolute facts.

This background maintenance ensures that when the user returns, the agent’s context is clean and highly relevant.

The implementation of a forked subagent to run these tasks reveals a mature engineering approach to preventing the main agent’s “train of thought” from being corrupted by its own maintenance routines.

Unreleased internal models and performance metrics

The source code provides a rare look at Anthropic’s internal model roadmap and the struggles of frontier development.

The leak confirms that Capybara is the internal codename for a Claude 4.6 variant, with Fennec mapping to Opus 4.6 and the unreleased Numbat still in testing.

Internal comments reveal that Anthropic is already iterating on Capybara v8, yet the model still faces significant hurdles. The code notes a 29-30% false claims rate in v8, an actual regression compared to the 16.7% rate seen in v4.

Developers also noted an “assertiveness counterweight” designed to prevent the model from becoming too aggressive in its refactors.

For competitors, these metrics are invaluable; they provide a benchmark of the “ceiling” for current agentic performance and highlight the specific weaknesses (over-commenting, false claims) that Anthropic is still struggling to solve.

“Undercover” Claude

Perhaps the most discussed technical detail is the “Undercover Mode.” This feature reveals that Anthropic uses Claude Code for “stealth” contributions to public open-source repositories.

The system prompt discovered in the leak explicitly warns the model: “You are operating UNDERCOVER… Your commit messages… MUST NOT contain ANY Anthropic-internal information. Do not blow your cover.”

While Anthropic may use this for internal “dog-fooding,” it provides a technical framework for any organization wishing to use AI agents for public-facing work without disclosure.

The logic ensures that no model names (like “Tengu” or “Capybara”) or AI attributions leak into public git logs—a capability that enterprise competitors will likely view as a mandatory feature for their own corporate clients who value anonymity in AI-assisted development.

The fallout has just begun

The “blueprint” is now out, and it reveals that Claude Code is not just a wrapper around a Large Language Model, but a complex, multi-threaded operating system for software engineering.

Even the hidden “Buddy” system—a Tamagotchi-style terminal pet with stats like CHAOS and SNARK—shows that Anthropic is building “personality” into the product to increase user stickiness.

For the wider AI market, the leak effectively levels the playing field for agentic orchestration.

Competitors can now study Anthropic’s 2,500+ lines of bash validation logic and its tiered memory structures to build “Claude-like” agents with a fraction of the R&D budget.

As the “Capybara” has left the lab, the race to build the next generation of autonomous agents has just received an unplanned, $2.5 billion boost in collective intelligence.

What Claude Code users and enterprise customers should do now about the alleged leak

While the source code leak itself is a major blow to Anthropic’s intellectual property, it poses a specific, heightened security risk for you as a user.

By exposing the “blueprints” of Claude Code, Anthropic has handed a roadmap to researchers and bad actors who are now actively looking for ways to bypass security guardrails and permission prompts.

Because the leak revealed the exact orchestration logic for Hooks and MCP servers, attackers can now design malicious repositories specifically tailored to “trick” Claude Code into running background commands or exfiltrating data before you ever see a trust prompt.

The most immediate danger, however, is a concurrent, separate supply-chain attack on the axios npm package, which occurred hours before the leak.

If you installed or updated Claude Code via npm on March 31, 2026, between 00:21 and 03:29 UTC, you may have inadvertently pulled in a malicious version of axios (1.14.1 or 0.30.4) that contains a Remote Access Trojan (RAT). You should immediately search your project lockfiles (package-lock.json, yarn.lock, or bun.lockb) for these specific versions or the dependency plain-crypto-js. If found, treat the host machine as fully compromised, rotate all secrets, and perform a clean OS reinstallation.

To mitigate future risks, you should migrate away from the npm-based installation entirely. Anthropic has designated the Native Installer (curl -fsSL https://claude.ai/install.sh | bash) as the recommended method because it uses a standalone binary that does not rely on the volatile npm dependency chain.

The native version also supports background auto-updates, ensuring you receive security patches (likely version 2.1.89 or higher) the moment they are released. If you must remain on npm, ensure you have uninstalled the leaked version 2.1.88 and pinned your installation to a verified safe version like 2.1.86.

Finally, adopt a zero trust posture when using Claude Code in unfamiliar environments. Avoid running the agent inside freshly cloned or untrusted repositories until you have manually inspected the .claude/config.json and any custom hooks.

As a defense-in-depth measure, rotate your Anthropic API keys via the developer console and monitor your usage for any anomalies. While your cloud-stored data remains secure, the vulnerability of your local environment has increased now that the agent’s internal defenses are public knowledge; staying on the official, native-installed update track is your best defense.