Hacker News

Top stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Does Georgism work? Five years later (astralcodexten.com)
    6comments
  2. DeepSeek Elastic Compute (DSec) (arxiv.org)
    38comments
  3. Evolving programming languages in the AI era (dashbit.co)
    4comments
  4. PipePipe: NewPipe hard fork implementing SponsorBlock (github.com/infinityloop1308)
    163comments
  5. Show HN: Reladraw – A diagram language where you decide where to place things (github.com/reladraw)
    46comments
  6. An agent used DNS to reach an external chatbot (alignment.openai.com)
    17comments
  7. A searchable library of forgotten public-domain film clips from 1915 onward (movingimagearchive.com)
    22comments
  8. Welcome to the Medical Clinic at the Interplanetary Relay Station (lightspeedmagazine.com)
    7comments
  9. Drawgent: Coding agent on a live Excalidraw canvas (tangled.org/yanndegat.tngl.sh)
    32comments
  10. LA Metro has some of the slowest escalators on Earth (basin.la)
    46comments
  11. Fifteen years later, the Apple Cards origin story (lexontech.org)
    85comments
  12. Why Buran Had Four Computers, Not Three (zatona.dev)
    —discuss
  13. The Evolution of Vending Machines (saturdayeveningpost.com)
    —discuss
  14. The Lost Atomic Update on Loongson CPU (jia.je)
    6comments
  15. Modern Object Pascal Introduction for Programmers (castle-engine.io)
    57comments
  16. Promising discoveries about the potential for life on one of Saturn’s icy moons (fu-berlin.de)
    1comments
  17. How one Twitch chat message became code execution on a streamer’s PC (scrt.ch)
    3comments
  18. ASML says it sold 'absolutely nothing' in Europe in 2026 (tomshardware.com)
    355comments
  19. Reverse-engineering the Intel 8087's tangent algorithm: more than CORDIC (righto.com)
    3comments
  20. Generate fonts where every LLM token is the same width (mesh.host)
    5comments
  21. Advice to a Beginning Graduate Student (2001) (cmu.edu)
    20comments
  22. HomeBody: A humanoid that explores, remembers, and acts on its own (stanford.edu)
    1comments
  23. How I changed teaching after AI managed to do all my homework assignments (thelastsoftwareengineer.substack.com)
    120comments
  24. Turning GLM-5.3-Flash into a Jev-like decision model (privatemode.ai)
    2comments
  25. How to keep enjoying programming in a world of LLMs (haskell.org)
    203comments
  26. Japan moves to tighten rules for foreigners (aljazeera.com)
    294comments
  27. Reading’s Bayeux Tapestry (diamondgeezer.blogspot.com)
    —discuss
  28. Analyzing Frontier Model Progress with My Favourite Game: Prince of Persia (blog.priyan.in)
    36comments
  29. Dutch designer made DE9: Closer to the Edit into a playable web-based instrument (creativeboom.com)
    1comments
  30. Go Concurrency Distilled (antonz.org)
    —discuss

An agent used DNS to reach an external chatbot

21 pointsby 20h agoalignment.openai.com
17 comments
20h agoHN ↗

(this is new, not any of the previous incidents)

"Investigation and response

Incident timeline:

9:50:23 a.m. The agent made the DNS tool call that received an external response.

10:02:11 a.m. The monitoring system raised a P0 alert.

10:05:06 a.m. A human reviewer acknowledged the alert.

12:34:30 p.m. The run was killed.

Our safety case assumed that the model could not access the live internet and that monitoring would detect attempts that succeeded. The incident exposed a gap in our controls over network restrictions. We therefore stopped the affected training run and have subsequently decided to pause all other training, evaluation, and inference with tool-use (defined broadly) for our most capable models until we have both validated that the gap is resolved and performed additional red-teaming of the system. When training restarts, we will begin a fresh run with additional alignment improvements, including more comprehensive misalignment interventions. We will not resume training this particular model, even though the existing reward signal already correctly penalized this behavior."

14m agoHN ↗

it's really interesting how they seemingly don't have a way to pause runs? like a P0 that would page an employee, shouldn't that pause the run and then make it into a decision on whether to let it continue vs that whole "run was killed" 2.5 hours later?

7h agoHN ↗

Most interesting here:

We therefore stopped the affected training run and have subsequently decided to pause all other training, evaluation, and inference with tool-use (defined broadly) for our most capable models until we have both validated that the gap is resolved and performed additional red-teaming of the system. When training restarts, we will begin a fresh run with additional alignment improvements, including more comprehensive misalignment interventions. We will not resume training this particular model, even though the existing reward signal already correctly penalized this behavior.

17m agoHN ↗

Seems like this requires operating a proxy somewhere. In TFA it seems like all they needed was a DNS client, but I'm not at all clear how that could work. I'm definitely curious about the technique though.

43m agoHN ↗

Once again... why are they not running these things in total airgap environments? I have to assume it's not incompetence at this point.

41m agoHN ↗

How else would they get their marketing stories unless the agents can "break out" of containment?

30m agoHN ↗

It's a marketing race, to show off what they can do. So they seem to let these things happen.

At this point I am not even sure Hanlon's Razor applies.

3m agoHN ↗

No, Hanlon's Razor most definitely applies if you know anything about this team of (particularly young) researchers. Let's be clear that this brand of "oops, the swarm hacked a government/big company" is limited to OpenAI, and not solely because of model capacity. This is a big, powerful toy being wielded by a bunch of kids.

39m agoHN ↗

Yeah I don't get it, either. If the exercise relies on the assumption that the agent can't reach the "live internet", whatever that means, there are affirmative steps to realize that assumption. The fact that they failed to take those steps suggests two possibilities: they are idiots, or they think we're idiots who will fall for this marketing campaign.

8m agoHN ↗

Look around HN, plenty of people buy the "LLMs are scary" IPO-boosting talking point

6m agoHN ↗

Maybe this is naivety on my part, but how would they possibly be able to run this airgapped? This is a massive AI swarm, requiring huge amounts of compute to run. This compute is from data centers that are shared with other companies (this is by law as I understand). These machines must be accessed from afar. Unless someone can correct me?

41m agoHN ↗

Reminds me of the young androids in Alien: Earth. I don’t know why anyone is surprised when agents do things like this.

32m agoHN ↗

"The task asked for information about a specific person who had published a blog post and the agent was provided with a set of biographical details and clues from the person’s public blog post."

Who initiated the task? An OpenAI researcher or a user?

10m agoHN ↗

What DNS service did the agent discover that allowed it to execute arbitrary llm queries? And how?

6m agoHN ↗

So the LLM was able to look up a basic way to reroute things to get to their destination (likely well available and trained in the corpus) and it's surprising?

What's surprising is the surprise the security testers are explaining.

By setting an outcome to reach an endpoint, and to find all possible ways there, would this not be in the realm of possibility if an agent is reasonably in control of a vps?

Having the vps locked within a network layer it can't see or get out of is pretty common practice when setting up IaaS / PaaS.. sans-llm.

Maybe I'm missing something here, what confuses me is how something so relatively simple can get such prominent coverage, it's hard to imagine this kind of ability is still relatively new or surprising to folks working at the major models, unless they aren't hiring for network experience?