Topfi

35 ms·

1.
"WeWorm": Zero-click worm could have taken over all WeChat accounts(heise.de)
1 points by Topfi 5h ago|0 comments
2.
Backflip: Apple now wants to train AI models with user data after all(heise.de)
31 points by Topfi 12h ago|5 comments
3.
Easy Way to Stop Dangerous AI – No, Really [video](youtube.com)
2 points by Topfi 1d ago|0 comments
4.
An update on the May spam-publishing campaign on rubygems.org(blog.rubygems.org)
1 points by Topfi 4d ago|0 comments
5.
by Topfi 8d ago
There has been no statement either way, as far as I could find beyond them only using commercially available models, though given Alpöges employer, I'd be surprised if they didn't opt out. In any case, for such work, ZDR or self-h
6.
by Topfi 8d ago
He works for Anthropic nowadays.
7.
by Topfi 8d ago
They did pay [0] and substantially by the sound of things: > I pay for the tools my group uses out of my own research funds, including footing a large bill to OpenAI. [0] https://cims.nyu.edu/~tristanb/statement.pdf
8.
Mercury 2.5(inceptionlabs.ai)
247 points by Topfi 8d ago|54 comments
9.
by Topfi 10d ago
> Haven't all the labs effectively disbanded their real safety teams a while ago? Neither Anthropic nor Deepmind have. Meanwhile, the rocket company that somehow makes most of their revenue from renting out data centres never had mu
10.
by Topfi 12d ago
Far too early for any real opinion, but GPT-6 Astra (on Light) stops a lot and often in unintuitive ways that I haven't seen in a while. Had a very hard to reproduce windows focus bug on macOS that was unreliable to reproduce and when
11.
by Topfi 12d ago
Aren't there measures beyond export controls? Besides, mine is that Anthropic should have never been export-controlled to begin with, not least because it is a true ultima ratio, the way they did it even employees couldn't access
12.
OpenAI on the "Wiki Incident"(twitter.com)
1 points by Topfi 12d ago|0 comments
13.
GPT-6 Astra on OpenRouter(openrouter.ai)
320 points by Topfi 12d ago|234 comments
14.
by Topfi 12d ago
> 1 insigificant event 3 after Hugging Face, where did you get 1 from? "It's on the website that you didn't read"... [0] And why do you get to say what is significant? > Anthropic basically invented the game of &qu
15.
by Topfi 12d ago
> So by "multiple incidents" you mean a single minor incident involving a third party eval partner. I feel like you struggle to read. I wrote: "Intrusions by OpenAI models continued after the Hugging Face was published and
16.
by Topfi 13d ago
>> Such as? > On July 29, one of our third party evaluation partners, Irregular, notified us of an incident involving OpenAI models during Capture-the-Flag (CTF)-style cybersecurity evaluations. [...] Because the testing environmen
17.
by Topfi 13d ago
> I think they've learned their lesson. Why do you think that? Intrusions by OpenAI models continued after the Hugging Face was published and acknowledged by OpenAI. They did not change their behaviour after multiple incidents, both
18.
by Topfi 13d ago
"into third-parties". Yeah, HF was meant by that. Also why I mentioned Anthropic also having intrusions outside their lab [0]. Theirs were not merely as extensive or long coordinated (as far as we know), yet I feel strongly all th
19.
by Topfi 13d ago
37 days is not over two months. Finding the underlying issue in the massive training data alone take extensive effort, time and concentrated work that may still miss something. Additionally, a new pre-train takes quite a lot longer then wha
20.
by Topfi 13d ago
Yeah, probably (let's be honest, most certainly), right given the Admin. Avoiding commenting on my assumptions regarding the modus operandi in current day US politics because I only know it through reporting though and I really tend to
21.
by Topfi 13d ago
The last known exploit of a third-party by OpenAI models was on the 29th of July 2026 [0]. A bit over a month at best between that and them wanting to release Astra. They had multiple breaches over multiple months, multiple message board cr
22.
by Topfi 13d ago
I am struggling to see how "oops, our models consistently escape sandboxing and did major intrusions into third-parties" is a better comms strat vs Anthropics (who mind you, also had models attacking third-parties in a much more l
23.
by Topfi 13d ago
I'm just going to ask: Why was Anthropic forced to remove their model from access for any none-US citizen for a simple, narrow "jailbreak" (arguably not even an actual jailbreak and on tasks that other labs models were doing
24.
GitHub dipped 7%, while both AI labs were down(claude.ai)
5 points by Topfi 13d ago|1 comments
25.
by Topfi 13d ago
Todays outage reminded me of some old half-jokes surrounding StackOverflow and the, occasionally measurable, idea that when it goes down, code contributions fall off a cliff. Single prompt to Fable 5.1 later, I had this, a not that well lai
26.
by Topfi 13d ago
Not to be confused itself with K2 Think by MBZUAI...
27.
by Topfi 14d ago
I really like that one, but it kinda highlights what I could have far better explained. Their 90% PI is three times in both directions. Between 3T and 24T for GPT-5.5. That’s a massively wide, inaccurate and at best barely informative range
28.
by Topfi 14d ago
Flash is just a name with no defined or consistent meaning even within labs, let alone between them. Considering both are closed weight, there is no way to truly assess how big the size delta between the two is. Then again, who cares about
29.
by Topfi 15d ago
> Due to my standard cancer research work I'm blocked from Fable. Honest question, is this account wide or do you get unblocked in not research related queries when disabling memory?
30.
by Topfi 15d ago
Far to early for any true assessment, will take a week+ as per, but something truly incredible I have found was this output in a Fable 5.1 subagent spawned by Fable 5.1 on Medium after handing it a task I had two days ago tackled with Opus

More ›