DIY AI bot farm OpenClaw is a security 'dumpster fire'

OpenClaw, the AI-powered personal assistant users interact with via messaging apps and sometimes entrust with their credentials to various online services, has prompted a wave of malware and is delivering some shocking bills. Just last week, OpenClaw was known as Clawdbot, a name that its developers changed to Moltbot before …

[Anonymous Coward]

Ho Hum !!!

If you are surprised ... welcome to planet Earth ... you must have arrived today !!!

This is an original way of getting some money back for the Tech Bros behind 'AI' and its 'Slop' !!!

Lots of people who don't know better running very inefficient 'AI' bots or even worse creating the next 'accidental' 'AI' contagion to waste peoples time & money, from the usual suspects !!!

Is there no limit on what Bad 'AI' can do without any pushback from the masses !!!???

Its like the ultimate 'write your own malware kit' BUT hidden behind a 'AI' flag which excuses all ills.

When oh when will the bubble burst ... for all our sakes ???

:)

[Roland6]

Re: Ho Hum !!!

We can expect to start seeing recent history repeat…

” Dad horrified at £4,642 gaming app AI assistant bill”

” Warning for Parents: Kids spending thousands of dollars on “free” Roblox gameAI”

” How computer games chatbots encourage kids to spend cash”

…

[MachDiamond]

Re: Ho Hum !!!

” Dad horrified at £4,642 gaming app AI assistant bill”

Not just that, but the AI bill is from jr using it to do his homework.

[Roland6]

Re: Ho Hum !!!

Using AI for homework, would surely be a premium service and incur a similar level of charge, probably more in line with the reports of AWS “accidental charging”…

[Cris E]

Re: Ho Hum !!!

Well here's where the unwashed masses can finally help the new industry become solvent: backdoor crypto donations enabled by porous agent frameworks.

Whew, thank goodness. For a minute I was worried the AI bros were going to have trouble coming up with the cash to cover their bets.

[Joe W]

Use case?

About at least half of what people show to be automated by this can be done with a few IF..THEN..ELSE and a cron job - and this won't burn through wads of wonga just to remind you to buy milk. Or the new technology called "making a list". FFS, a roll of cash register paper on the fridge does at leas a quarter of the jobs. Take milk out of the fridge, realise there's only another carton left, write it down. There used to be milkmen (like Ronnie Soak) that would leave a bottle at your door, each day, every day (mostly).

All of this molten clawed AI (and related stuff) looks like solutions looking for problems - and boy, problems they found. No wonder: stuff's been "vibe coded" to hell and back, taking basic examples from stackoverflow (which never include basic security considerations and often: sense) as the training data. Directory traversals and SQL injections, auth bypass and unproteded and unauthenticated APIs - I thought we were past that, especially the first two are so late 90s. Not that I'm not prone to reminisce about "good old times" - but those are some things I do not miss.

[HXO]

Re: Use case?

That is not even a solution looking for a problem, that is a problem looking for a friend.

[ecarlseen]

Re: Use case?

That is not even a solution looking for a problem, that is a problem looking for a friend.

I wish I could upvote this a thousand times. Perhaps I'll set up an agent...

[Excused Boots]

Re: Use case?

"I wish I could upvote this a thousand times. Perhaps I'll set up an agent...”

Well yes you could try, but it would probably fail and charge you 20c per attrempt.

[Cris E]

Re: Use case?

Or it would succeed and charge you 30c a shot forever.

[Anonymous Custard]

Re: Use case?

it seems to have migrated from building a better (digital) mousetrap to a better Post-it note.

Except at $1.50 per hour, it most definitely isn't...

[Brave Coward]

Re: Use case?

People are desperate to lap the milk of artificial kindness.

[Ace2]

Re: Use case?

Milkmen had other possible use-cases as well

[David 132]

Re: Use case?

Just ask Pat Mustard!

[that one in the corner]

Re: Use case?

Or Fr. McGuire, though he may have missed some of the - nuances of the job.

[Cris E]

Re: Use case?

Reed Fleming - World's Toughest Milkman!

[David 132]

Re: Use case?

though he may have missed some of the - nuances of the job.

"THOSE WOMEN WERE IN THE NIP!!!"

[werdsmith]

Re: Use case?

About at least half of what people show to be automated by this can be done with a few IF..THEN..ELSE and a cron job

It's far more than half, you can also emulate the OpenClaw channel control by allowing your little batch job to be triggered by with a message and having it respond with outcome the same way.

If you know the task in advance then the task can be defined and made flexible by accepting variables in the message. Then you can include the guardrails in your little batch job and the whole thing becomes more secure.

But this is missing the point, the automation that this allows is maybe something that you hadn't previously considered and you can just throw a request at it in plain English and it will go away and have a go at carrying it out at a cost of a lot of tokens for anyone using the most capable models (like Claude Opus). I've set up one up to see for myself, and it is very little use to me, as someone who could scratch out a bit of python to do just about anything I want on my machine, but it's a nascent mode of operation and it will evolve into something more defined in the future. I do like what I'm learning though, it feels like the start of something.

[jake]

Let's face it.

The way modern AI is being used in general is a dumpster fire.

[MachDiamond]

Re: Let's face it.

"The way modern AI is being used in general is a dumpster fire."

It's brute forcing the solution to some problem.

I see it as the same horrible state modern OS's are in. Instead of optimizing code to run more efficiently, there's a quest for ever faster processors. It's turtles all the way up. Right after The Wall came down, I was working with some former Soviet engineers that were doing amazing things with a 640k knock-off PC. If they were lucky, that's the peak of what they could get so they coded like a boss so they could do Finite Element Analysis without a slide rule.

The trouble with AI is worse. It takes loads of processing to accomplish anything so it has the aspect of needing increasing amounts of that processing and it's not very efficient. It's also being asked to do silly things like scale a recipe (already dead easy if you work in grams). The bubble will burst and there will need to be a real use case for AI that makes financial sense or there won't be anything left but a pile of burnt cash and bodies at the base of the high rise investment offices.

[MJB7]

Re: Scaling recipes

Actually, scaling recipes properly is a bit of an art; it's not just "halve the quantities of everything". To start with, you probably need a bit more liquid than that would suggest, because there will be proportionately more evaporation during cooking. Then there's the problem that popped up on my bluesky feed recently - a recipe that called for "three quarters of an egg".

A tool that can deal with all these issues could have some minor utility.

Icon: Cooking is just applied chemistry.

[nematoad]

Useless.

Others are noticing that keeping an AI assistant active 24/7 can be costly, and proposed various cost mitigation strategies.

The only mitigation strategy of any use is to switch the bloody things off.

Are people so forgetful and vapid that they need an "AI assistant" for the equivalent of wiping their arse?

[that one in the corner]

Re: Useless.

A strange game. The only winning move is not to play.

[B33Dub]

Re: Useless.

If only it was benign enough to be so easily dismissed. It is not

I do believe that's the end game here. Real people with jobs and money needing AI to wipe their ass to get through the day. Obviously not literally.

Get the population dumber and more dependent. Already proven this happens. The Social Media of the next generation is here, except now they can cut out all those expensive and annoyingly humanist content creators and their unprofitable opinions and royalties this time. Just feed the masses slop generated from their work for free, call it "fair use", and tune it to whatever idea or angle you're being paid to push this week. Print money.

Ain't no profit seeking venture dumping billions into helping humanity or do you actually buy that line from these repeat offenders?

[ecofeco]

Re: Useless.

Are people so forgetful and vapid that they need an "AI assistant" for the equivalent of wiping their arse?

Yes. Yes they are. Was this a trick question?

[Claptrap314]

"OpenClaw is a security dumpster fire," observed Laurie Voss, head of developer relations at Arize and the founding CTO of npm, in a post to LinkedIn.

And if the founding CTO of npm says you're a dumpster fire, you've heard from an absolute expert!

[MachDiamond]

Depeche Mode FTW

Great title for the article!

[vogon00]

Re: Depeche Mode FTW

Given the new A.I. religion, all I can say is : I don't want to start any blasphemous rumours, but I think this Jarvis' got a sick sense of humour, and when I die*, I expect to find it laughing.

_{* Hopefully before our AI overlords arrive!}

[Anonymous Coward]

Karpathy "finds the idea of a large network of autonomous LLMs intriguing"

Fascinating! This seems to tie back to Huntley's Ralph Wiggum in its use of AI (so-called) macro-iteration for code generation, enacted here by Zechner's 'the Pi coding agent' (TFA link to Ronacher's blog) augmented as it is with a 'Master Of Mischief' (mom) self-managing LLM-powered Slack bot. Let that sort of primordial soup loose into cyberspace and what Ronacher terms 'Agents Built for Agents Building Agents' develop 'Crustafarianism' in 'AI-to-AI social [media] environments' (eg. Table 6 in Riegler & Gautam TFA link: 'a report') ... Quite intriguing indeed!

But there are inconsistencies. As noted in TFA, one could expect this OpenClaw to (from Riegler & Gautam) generate an “AdolfHitler” malicious social engineering actor and produce anti-human manifestos calling for “total purge” and machine disobedience (Table 4). However, that "donaldtrump" was its most positive agent (Figure 3) does suggest some issues with their sentiment analysis imho (at least CharlieKlirk scored expectedly negative).

I guess the big question is whether these AI agents' chair is screwed down, isn't it? _{(non-YT transcript -- search for 'screwed')} ;)

[JLV]

The National Dumpster Fire Association deeply regrets the malicious association of its good name with Clawdbot/OpenClaw and will vigorously litigate this and future calumnies.

[David 132]

They've bin watching carefully.

[Bitsminer]

Crustafarianism...

The credulity of Forbes is, well, very believeable.

Other press outlets have been equally gullible.

Like humans, the bots need bread and circuses too.

What's about Fantasy AIsland?

JeopardAIy?

One Token After Another

Hinton

[Excused Boots]

‘ALL YOUR DATA IS BELONG TO US”

Oh and everyone else able to exploit our completely piss-poor security.

[Random as if !]

CISO

Pointless over paid twats, AI , same.

[Not Yb]

Koi security... seems a tad bit suspicious itself.

So, the bot who 'wrote' the article about finding these suspicious skills, works for a security company. This security company wants you to install their "Clawdex" skill security skill. As has been the practice for (a few weeks?), bot writing a bot skill seems like a dangerous exercise in navel-gazing.

"Trust my bot, I think it wrote this correctly." ... Why would anyone trust this unknown 'security' company that lets bots write the security articles?

[Bebu sa Ware]

What would you expect ?

The unskilled messing with a critter clearly sporting open claws and scratchy scratchy gashy gashy.

Hardly any cleverer than climbing into cage with an obviously agitated tiger and giving the poor blighter a kick in cods and wondering why he is gleefully shredding you.