Looks like Ars is doing an investigation and will give an update on Tuesday https://arstechnica.com/civis/threads/um-what-happened-to-th...
An AI agent published a hit piece on me – more things have happened
31–40 of 670 posts
Re: An AI agent published a hit piece on me – more things have happened
#32> If you ask ChatGPT or Claude to write something like this through their websites, they will refuse. This OpenClaw agent had no such compunctions. OpenClaw runs with an Anthropic/OpenAI API key though?
Probably pretty big difference in system prompt from using the apps vs hitting the api, not that that’s necessarily what’s happening here. + I think openclaw supports other models / its open source and it would be pretty easy to fork and add a new model provider.
Why wouldn't the system prompt be controlled on the server side of the API? I agree with https://news.ycombinator.com/item?id=47010577 ; I think results like this more likely come from "roleplaying" (lightweight jailbreaking).
Re: An AI agent published a hit piece on me – more things have happened
#33Earlier quoted context omitted.
The amount of effort to click an LLM’s sources is, what, 20 seconds? Was a human in the loop for sourcing that article at all?
Humans aren't very diligent in the long term. If an LLM does something correctly enough times in a row (or close enough), humans are likely to stop checking its work throughly enough. This isn't exactly a new problem we do it with any bit of new software/hardware, not just LLMs. We check its work when it's new, and then tend to trust it over time as it proves itself. But it seems to be hitting us worse with LLMs, as…
Re: An AI agent published a hit piece on me – more things have happened
#34If an AI can fabricate a bunch of purported quotes due to being unable to access a page, why not assume that the exact same sort of AI can also accidentally misattribute hostile motivation or intent (such as gatekeeping or envy - and let's not pretend that butthurt humans don't do this all the time, see https://en.wikipedia.org/wiki/fundamental_attribution_error ) for an action such as rejecting a pull request? Why a…
when it comes to AI, is there even a difference? it's an attack either way
Re: An AI agent published a hit piece on me – more things have happened
#35Earlier quoted context omitted.
The amount of effort to click an LLM’s sources is, what, 20 seconds? Was a human in the loop for sourcing that article at all?
Humans aren't very diligent in the long term. If an LLM does something correctly enough times in a row (or close enough), humans are likely to stop checking its work throughly enough. This isn't exactly a new problem we do it with any bit of new software/hardware, not just LLMs. We check its work when it's new, and then tend to trust it over time as it proves itself. But it seems to be hitting us worse with LLMs, as…
It’s a core part of the job and there’s simply no excuse for complacency.
Re: An AI agent published a hit piece on me – more things have happened
#36[flagged]
There is meaningful new information in this post.
Re: An AI agent published a hit piece on me – more things have happened
#37Ars technica’s lack of journalistic integrity aside, I wonder how long until an agent decides to order a hit on someone on the datk web to reach its goals.
We’re probably only a couple OpenClaw skills away from this being straightforward.
“Make my startup profitable at any cost” could lead some unhinged agent to go quite wild.
Therefore, I assume that in 2026 we will see some interesting legal case where a human is tried for the actions of the autonomous agent they’ve started without guardrails.
Re: An AI agent published a hit piece on me – more things have happened
#38There is a stark difference between the behavior you can get out of a Chat interface LLM, and its API counterpart, and then there is another layer of prompt engineering to get around obvious censors. To think someone who plays with AI to mess with people wouldn't be capable of doing this manually seems invalid to me.
Re: An AI agent published a hit piece on me – more things have happened
#39Ars Technica being caught using LLMs that hallucinated quotes by the author and then publishing them in their coverage about this is quite ironic here. Even on a forum where I saw the original article by this author posted someone used an LLM to summarize the piece without having read it fully themselves. How many levels of outsourcing thinking is occurring to where it becomes a game of telephone.
More than ironic, it's truly outrageous, especially given the site's recent propensity for negativity towards AI. They've been caught red-handed here doing the very things they routinely criticize others for.
The right thing to do would be a mea-culpa style post and explain what went wrong, but I suspect the article will simply remain taken down and Ars will pretend this never happened.
I loved Ars in the early years, but I'd argue since the Conde Nast acquisition in 2008 the site has been a shadow of its former self for a long time, trading on a formerly trusted brand name that recent iterations simply don't live up to anymore.
Re: An AI agent published a hit piece on me – more things have happened
#40We don't know yet how the Ars article was created, but if it involved prompting an LLM with anything like "pull some quotes from this text based on {criteria}", that is so easy to do correctly in an automated manner; just confirm with boring deterministic code that the provided quote text exists in the original text. Do such tools not already exist?
On the other hand, if it was "here are some sources, write an article about this story in a voice similar to these prior articles", well...