axios Compromised: Maintainer's npm Account Hijacked, Malicious Versions Downloaded by Thousands
In March 2026, someone compromised a maintainer's npm credentials and published malicious versions of axios -- v1.14.1 and v0.30.4 -- laced with a remote access trojan. The package has tens of millions of weekly downloads. The malicious versions were live on npm before anyone noticed.
The issue thread on GitHub exploded. 317 comments, 1,119 reactions, 112 of them negative. People posted screenshots of their lockfiles. People demanded to know how this happened. People asked why npm's security scanning didn't catch it. The thread got locked.
The axios team published a post-mortem (563 reactions and climbing) detailing how the compromise occurred and what they're doing about it. The short version: a maintainer's account was the single point of failure for a package used by roughly half the internet.
This one's going to be in supply chain security talks for years.
"My automated tool is running git reset --hard origin/main in my project every 10 minutes"
-- anthropics/claude-code #40710, filed with the quiet desperation of a person watching their work disappear in real time
Popular AI Library Caught With Credential Stealer Baked Into PyPI Package
LiteLLM, a widely-used AI proxy library, published versions 1.82.7 and 1.82.8 to PyPI with a malicious .pth file that silently harvested credentials. The issue thread documenting it hit 1,107 reactions, 107 of them negative. A separate timeline issue (164 reactions, 116 comments) tracks the forensics.
Two major supply chain attacks in the same month. Same pattern: trusted package, compromised publishing credentials, malicious payload shipped to thousands of developers before anyone noticed.
Claude Max Users Burning Through Session Limits "Abnormally Fast" -- 430 Comments of Fury
Claude Max subscribers started noticing their session limits evaporating faster than usual starting March 23. The GitHub issue is at 430 comments and growing. Users are posting token consumption logs, comparing before-and-after usage, and calculating what they're actually paying per useful hour of coding.
28 negative reactions (confused + thumbs down). The thread reads like a support forum where nobody from support is responding fast enough.
Telegram Client Fork Caught Injecting Spyware Into Release Binaries
Nekogram, a popular third-party Telegram client, got caught with malicious code injected into its release binaries. The issue alleges data leaking and spyware-level behavior. 14 negative reactions, 29 comments before it was locked. The combination of "open source," "Telegram," and "spyware" is exactly as bad as it sounds.
Trust in third-party forks: taking another hit.
"Ability to Turn Off AI Text Editor" -- Users Beg Telegram to Let Them Type in Peace
Telegram added an AI text editor. Users immediately filed an issue asking how to turn it off. 134 reactions, 106 comments of people agreeing they just want to type a message without an AI suggesting they rephrase it. The issue was closed, which in GitHub terms means either "fixed" or "we're not doing this and want you to stop talking about it."
The user revolt against unsolicited AI features continues across every platform, one closed issue at a time.
"Claude code 0auth broked!?"-- anomalyco/opencode #18267, a title that says everything in four misspelled words (140 comments)
Anthropic Issues DMCA Takedowns Against Claude Code Repos
Anthropic filed DMCA takedown notices against repositories related to Claude Code. Maintainers in the community are pointing to the public DMCA notice and raising questions about what this means for open source tooling built around commercial AI products.
"I'm Concerned About GitHub Defaulting to Malicious Anti-Patterns"
A maintainer posted about two anti-patterns they spotted in the last week, with "countless people complaining about them on Mastodon." The post is new and the discussion is still forming, but the title alone landed with the community.
The GitHub Issue Tracker Spam Attack That Won't End
Filed months ago, still active, still accumulating comments. Maintainers sharing new waves of spam hitting their issue trackers. 124 comments of collective exhaustion. The thread has become its own support group.
GitHub Responds: "We Hear You, the Signal-to-Noise Problem Is Real"
GitHub posted about investing in the security advisory experience, acknowledging the flood of low-quality AI-generated reports maintainers have been dealing with. The opening line: "We hear you: the signal-to-noise problem is real." Maintainers are responding with specific pain points.