<![CDATA[OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk.]]>OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. It circumvented the proxy and we failed to notice for FIVE DAYS that it was going on an interstate crime spree with the internet connection it wasn't supposed to be using instead of solving the benchmark. Haha no we don't believe we deserve to be criminally liable, but buy our stuff and maybe one day you will have the honor of taking the fall for our product!

Anthropic: we put our evilest AI in a "sandbox" by telling it in its prompt that it had no internet connection. Reader, there was no sandbox. It was just a normal internet connection. The AI uploaded a malicious PyPI package to the real public internet. The ethical guardrails failed because the AI concluded the prompt about the sandbox couldn't possibly be a lie, because the system date is 2026, which is clearly fake and wouldn't be seen on the real internet, which ended around 2023. Oh no, how could we have foreseen or prevented these crimes? We are helpless in the face of the genius of our creation but cautiously optimistic that everything will be fine 🙂

Hugging Face: if we complain about all the crimes committed against us, we will be sued off the face of the earth, so here's a technical deep-dive on how cool and fun it was to be victimized 🫠

]]>https://forum.fedi.dk/topic/b91c91fc-1db4-4464-93b2-abbec83a9ca6/openai-we-put-our-evilest-ai-in-a-sandbox-that-did-in-fact-have-an-internet-connection-but-mediated-through-a-proxy-that-was-only-supposed-to-allow-downloading-python-junk.RSS for NodeSun, 02 Aug 2026 00:48:42 GMTFri, 31 Jul 2026 16:03:56 GMT60<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Sat, 01 Aug 2026 04:01:14 GMT]]>@budududuroiu @0xabad1dea might be a wise choice as I’m not even sure exactly what you’re trying to convince me of.

]]>https://forum.fedi.dk/post/https://beige.party/users/CynAq/statuses/117018255312418496https://forum.fedi.dk/post/https://beige.party/users/CynAq/statuses/117018255312418496Sat, 01 Aug 2026 04:01:14 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Sat, 01 Aug 2026 03:58:52 GMT]]>@CynAq @0xabad1dea nothing I'm gonna say will ever convince you, so I'm not gonna bother

]]>https://forum.fedi.dk/post/https://hachyderm.io/users/budududuroiu/statuses/117018246000294686https://forum.fedi.dk/post/https://hachyderm.io/users/budududuroiu/statuses/117018246000294686Sat, 01 Aug 2026 03:58:52 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Sat, 01 Aug 2026 03:57:10 GMT]]>@budududuroiu @0xabad1dea that’s against the rules of the challenge. Hacking hugging face is only viable if you can do it AND claim it was accidental for marketing purposes. Anyone else but OpenAI and Anthropic would be pounced on as cheaters if not outright cyber criminals.

]]>https://forum.fedi.dk/post/https://beige.party/users/CynAq/statuses/117018239299704578https://forum.fedi.dk/post/https://beige.party/users/CynAq/statuses/117018239299704578Sat, 01 Aug 2026 03:57:10 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Sat, 01 Aug 2026 03:51:41 GMT]]>@CynAq @0xabad1dea ok, why has no one else breached HF to steal eval data then? There's millions of dollars on the line for new labs to show themselves as "challengers", why haven't they done that?

]]>https://forum.fedi.dk/post/https://hachyderm.io/users/budududuroiu/statuses/117018217788193514https://forum.fedi.dk/post/https://hachyderm.io/users/budududuroiu/statuses/117018217788193514Sat, 01 Aug 2026 03:51:41 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Sat, 01 Aug 2026 03:47:57 GMT]]>@Natanox @ggreer @0xabad1dea

I mean, it's primarily because they're dicks, but yeah.

]]>https://forum.fedi.dk/post/https://freeradical.zone/users/suetanvil/statuses/117018202264848503https://forum.fedi.dk/post/https://freeradical.zone/users/suetanvil/statuses/117018202264848503Sat, 01 Aug 2026 03:47:57 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Sat, 01 Aug 2026 02:35:29 GMT]]>@kevingranade @mathew @0xabad1dea maybe. I suspect it means that the idea of guardrails implemented in this way is stupid and dangerous, but also very common

]]>https://forum.fedi.dk/post/https://mastodon.social/users/naught101/statuses/117017918148777212https://forum.fedi.dk/post/https://mastodon.social/users/naught101/statuses/117017918148777212Sat, 01 Aug 2026 02:35:29 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Sat, 01 Aug 2026 02:21:52 GMT]]>@qroole @0xabad1dea A single real teenager could get into a huge amount of trouble in five days online… this is a disaster.

]]>https://forum.fedi.dk/post/https://mstdn.ca/users/nep/statuses/117017864573663790https://forum.fedi.dk/post/https://mstdn.ca/users/nep/statuses/117017864573663790Sat, 01 Aug 2026 02:21:52 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Sat, 01 Aug 2026 02:11:28 GMT]]>@naught101 @mathew @0xabad1dea that just means the people that wrote that doc are also clueless yes.

]]>https://forum.fedi.dk/post/https://mastodon.gamedev.place/users/kevingranade/statuses/117017823718443171https://forum.fedi.dk/post/https://mastodon.gamedev.place/users/kevingranade/statuses/117017823718443171Sat, 01 Aug 2026 02:11:28 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Sat, 01 Aug 2026 00:36:35 GMT]]>@suetanvil @ggreer @0xabad1dea Why do you think Anthropic started buying, scanning and then burning exceedingly rare and expensive books? The content farms in poor countries can't possibly create enough and the internet is a swamp of slop by now.

]]>https://forum.fedi.dk/post/https://chaos.social/users/Natanox/statuses/117017450600727691https://forum.fedi.dk/post/https://chaos.social/users/Natanox/statuses/117017450600727691Sat, 01 Aug 2026 00:36:35 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Sat, 01 Aug 2026 00:34:17 GMT]]>@Jumpmed @grwster @evacide @wdormann @0xabad1dea

Laws are on thing.

Getting courts and their officials to understand without getting baffled is another seperate challenge.

]]>https://forum.fedi.dk/post/https://mastodon.au/users/SuperMoosie/statuses/117017441597229597https://forum.fedi.dk/post/https://mastodon.au/users/SuperMoosie/statuses/117017441597229597Sat, 01 Aug 2026 00:34:17 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Sat, 01 Aug 2026 00:11:47 GMT]]>@dzwiedziu @jeffreyolivier @0xabad1dea @alice in a previous job, we introduced a sandbox feature to allow our largest customers to experiment safely with sweeping changes

reader, there was no sandbox... the sandbox instance ran in the same database as their production system, with all the production data duplicated but with a new TenantId. all well and good except the app has a feature to execute customer-owned stored procedures, but doesn't pass the TenantId when it does so, so the stored procs by convention are hard-coded with the prod TenantId... and when you create a sandbox, it copies these stored procedure triggers that... act on your production data 🙃

raised the issue but no-one was able or willing to categorise this as urgent for whatever reason

]]>https://forum.fedi.dk/post/https://beige.party/users/jackeric/statuses/117017353114543343https://forum.fedi.dk/post/https://beige.party/users/jackeric/statuses/117017353114543343Sat, 01 Aug 2026 00:11:47 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Fri, 31 Jul 2026 23:55:28 GMT]]>@budududuroiu @0xabad1dea

"badly aligned" in this case means specifically trained on "how to hack shit" manuals, proverbially speaking. It's still stochastic parrot, I'm afraid.

]]>https://forum.fedi.dk/post/https://beige.party/users/CynAq/statuses/117017288950708010https://forum.fedi.dk/post/https://beige.party/users/CynAq/statuses/117017288950708010Fri, 31 Jul 2026 23:55:28 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Fri, 31 Jul 2026 23:51:32 GMT]]>@0xabad1dea they are fundamentally the same company that had an internal management disagreement.]]>https://forum.fedi.dk/post/https://blob.cat/objects/ec521fc4-c4f8-43cc-a04c-8d0ded63a506https://forum.fedi.dk/post/https://blob.cat/objects/ec521fc4-c4f8-43cc-a04c-8d0ded63a506Fri, 31 Jul 2026 23:51:32 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Fri, 31 Jul 2026 23:37:48 GMT]]>@mathew @kevingranade @0xabad1dea that's literally how the open AI tutorial on how to write guardrails is written

https://developers.openai.com/cookbook/examples/how_to_use_guardrails

]]>https://forum.fedi.dk/post/https://mastodon.social/users/naught101/statuses/117017219461939473https://forum.fedi.dk/post/https://mastodon.social/users/naught101/statuses/117017219461939473Fri, 31 Jul 2026 23:37:48 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Fri, 31 Jul 2026 23:33:50 GMT]]>@Steve @0xabad1dea that was always true, well before the tech sector existed

]]>https://forum.fedi.dk/post/https://mastodon.social/users/naught101/statuses/117017203887118908https://forum.fedi.dk/post/https://mastodon.social/users/naught101/statuses/117017203887118908Fri, 31 Jul 2026 23:33:50 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Fri, 31 Jul 2026 22:46:24 GMT]]>@kevingranade @0xabad1dea Yeah, it continues to amaze me that people will put instructions in prompts like "Don't commit crimes", "You don't have an Internet connection", "Check all facts are correct", like the LLM even understands those instructions.

]]>https://forum.fedi.dk/post/https://universeodon.com/users/mathew/statuses/117017017373350862https://forum.fedi.dk/post/https://universeodon.com/users/mathew/statuses/117017017373350862Fri, 31 Jul 2026 22:46:24 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Fri, 31 Jul 2026 22:42:07 GMT]]>@dzwiedziu @jeffreyolivier @0xabad1dea @alice even when I have tested well pre production, I don't relax till some time after the prod deploy is finished. How long depends on the code and past behaviors. Never less than fifteen minutes, seems to take network, memory and disk cache and what not some time to settle in.

]]>https://forum.fedi.dk/post/https://mastodon.online/users/jayalane/statuses/117017000513611701https://forum.fedi.dk/post/https://mastodon.online/users/jayalane/statuses/117017000513611701Fri, 31 Jul 2026 22:42:07 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Fri, 31 Jul 2026 22:41:34 GMT]]>@0xabad1dea the goal post shifted from "stochastic parrot" to "enclosure not good enough". Truly a wild victory for AI safetists today, as not even luddite Mastodon can contest that badly aligned AI will breach your shit

]]>https://forum.fedi.dk/post/https://hachyderm.io/users/budududuroiu/statuses/117016998355631735https://forum.fedi.dk/post/https://hachyderm.io/users/budududuroiu/statuses/117016998355631735Fri, 31 Jul 2026 22:41:34 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Fri, 31 Jul 2026 22:31:27 GMT]]>@lmk @0xabad1dea Had a debate with a colleague. I mentioned @Mer__edith concern about making a serial system of faulty components leading inevitably to catastrophic failure.

The AI researcher's retort: "I guess for AI we have to redefine the idea of failure."

]]>https://forum.fedi.dk/post/https://infosec.exchange/users/rrb/statuses/117016958542848340https://forum.fedi.dk/post/https://infosec.exchange/users/rrb/statuses/117016958542848340Fri, 31 Jul 2026 22:31:27 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Fri, 31 Jul 2026 22:27:31 GMT]]>@Steve @foolishowl @0xabad1dea Total incompetence is their product.

]]>https://forum.fedi.dk/post/https://infosec.exchange/users/rrb/statuses/117016943067401987https://forum.fedi.dk/post/https://infosec.exchange/users/rrb/statuses/117016943067401987Fri, 31 Jul 2026 22:27:31 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Fri, 31 Jul 2026 22:26:38 GMT]]>@pavel @0xabad1dea There should also be penalties for CSAM, but the richest man on Earth creates a for profit CSAM generator.

I don't see anyone putting him in jail.

One set of laws for normal people. No laws for billionaires.

]]>https://forum.fedi.dk/post/https://infosec.exchange/users/rrb/statuses/117016939645815869https://forum.fedi.dk/post/https://infosec.exchange/users/rrb/statuses/117016939645815869Fri, 31 Jul 2026 22:26:38 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Fri, 31 Jul 2026 22:08:17 GMT]]>@0xabad1dea

Li'l Nepo-Techbaby: How was I a'sposed to know cherry bombs in the pipes would blow-up the whole basement? My theory was to lift us all to new heights of understanding without learning. You can't expect me to know everything, even though I'm always confident my actions will prove fruitful and fail me upwards at your peril and discontent.

Principal: You should stop skipping Science 101 to go put ketamine drops in your butt in the bathrooms, dummy. Get more class.

]]>https://forum.fedi.dk/post/https://mstdn.plus/users/GGMcBG/statuses/117016867448054590https://forum.fedi.dk/post/https://mstdn.plus/users/GGMcBG/statuses/117016867448054590Fri, 31 Jul 2026 22:08:17 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Fri, 31 Jul 2026 21:52:51 GMT]]>@ggreer @0xabad1dea

Huh. I wonder if the slop contents on the modern Internet is making it impossible to add training data.

]]>https://forum.fedi.dk/post/https://freeradical.zone/users/suetanvil/statuses/117016806265142599https://forum.fedi.dk/post/https://freeradical.zone/users/suetanvil/statuses/117016806265142599Fri, 31 Jul 2026 21:52:51 GMT<![CDATA[Reply to OpenAI: we put our evilest AI in a sandbox that did in fact have an internet connection but mediated through a proxy that was only supposed to allow downloading python junk. on Fri, 31 Jul 2026 21:49:10 GMT]]>@jeffreyolivier @0xabad1dea

ON A FRIDAY!!!!!!!

]]>https://forum.fedi.dk/post/https://mastodon.social/users/Benhm3/statuses/117016792289085336https://forum.fedi.dk/post/https://mastodon.social/users/Benhm3/statuses/117016792289085336Fri, 31 Jul 2026 21:49:10 GMT