neuronexmachina

Link to "National Security Memorandum-7" for reference: https://www.whitehouse.gov/presidential-actions/2025/09/coun...

Example quote:

>(h) The Attorney General shall issue specific guidance that ensures domestic terrorism priorities include politically motivated terrorist acts such as organized doxing campaigns, swatting, rioting, looting, trespass, assault, destruction of property, threats of violence, and civil disorder. This guidance shall also include an identification of any behaviors, fact patterns, recurrent motivations, or other indicia common to organizations and entities that coordinate these acts in order to direct efforts to identify and prevent potential violent activity.

Also worth noting that the main touted difference with Claude Mythos isn't it's ability to find vulnerabilities, but rather chaining them together to create full useable exploits. I haven't heard of any evidence that the Claude Fable "fix this code" jailbreak could have been used to do exploit-chaining.

Is there any reason they couldn't also apply export-control to older models, just to screw with Anthropic?

From reading the post, I think it's more likely that anti-jailbreaking is going to become much more strict and prone to false-positives.

> We received the directive from the government today at 5:21pm (ET). The letter did not provide specific details of its national security concern. Our understanding is that the government believes it has become aware of a method of bypassing, or “jailbreaking” Fable 5. We reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities. These vulnerabilities all appear relatively simple, and we have found that other publicly-available models are able to discover them as well without requiring a bypass.

Based on this, it seems like the Trump admin would have targeted them even without the "scaremongering":

> To date, the government has only given us verbal evidence of a potential narrow, non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws. Our understanding is that one potential jailbreak was shared with the government. We have reviewed the report and validated that the level of capability displayed there is widely available from other models (including OpenAI’s GPT-5.5), and is used every day by the defenders who keep systems safe.

The Texas Tribune is generally regarded as a reputable news source. Which quote are you referring to?

For GPT-2 and GPT-3 it seems like the concern was that they hadn't yet figured out how to properly write safeguards for it yet:

> The company believes making its API generally available was made possible due to its progress with safeguards, and that opening up the API to all developers will help see applications developed faster. ...

> A large emphasis has been placed on safe use of the tool, which in the past has been criticised for a range of shortcomings, including racism and prejudices against specific genders and religions.

From your earlier comment it sounded like there was a "heist" simply based on having a similar name. Looking into it though, it seems like the ARDC non-profit did a pretty reasonable job of proving they were the same folks who'd been managing the IP block for decades. Also, has there been any sort of allegation that they've misused the funds? From what I understand they've pretty consistently used the funds to support amateur radio.

Do you have a source for your claims about the ARDC?

Was curious and astral put up their AI policy a couple weeks before the acquisition. Of course, it's quite possible they already knew it was happening: https://github.com/astral-sh/.github/pull/1

That said, I'm kind of surprised ripgrep hasn't been acquired by anyone, considering all the major AI agents use it pretty heavily.

> If AI labs saw a lot of potential there, they'd surely be bragging about it non-stop?

Google seems to pretty regularly post about how their TPU and algorithm advancements have been decreasing energy costs for both inference and training.

TIL: https://en.wikipedia.org/wiki/Rieko_Kodama

> A memorial message dedicated to Kodama was included in the end credits of the Sega Genesis Mini 2, which was released on October 27, 2022.[16] Addressing inquiries by IGN and Famitsu, Sega confirmed that Kodama had died on May 9, 2022, aged 58,[17][18] and initially refrained from a public announcement to respect her family's privacy

CVEs:

* https://nvd.nist.gov/vuln/detail/CVE-2026-28952

* https://nvd.nist.gov/vuln/detail/CVE-2026-28942

> keeping the external costs (congestion, danger, pollution, noise, etc)

Are there actually self-driving car services that aren't EV-based? Just that helps with several of those criteria, and they're substantially less dangerous than human-driven vehicles.

Do you have a reference for that? At least based on this, it seems like China's trying to increase collaboration and funding for joint research projects with non-Chinese researchers:

* https://www.nsfc.gov.cn/english/site_1/international/D2/2018...

* https://www.france24.com/en/live-news/20260107-overseas-scho...

Can you elaborate? That just seems to be a Cloudflare's announcement from May 2025 that they'd be supporting MCP servers.

Altadena is part of unincorporated LA County, they don't vote in the City of Los Angeles elections.

The design files and firmware are open source, so you hypothetically produce your own: https://github.com/ploopyco/bean-pointing-stick

Which part was defensive?

Since this has a foreign-made processor and WiFi module, would this be blocked by the Trump FCC's foreign-made router ban?

I think it's essentially this plugin? https://github.com/anthropics/knowledge-work-plugins/tree/ma...

> So is this thing talking to Google's servers all the time for the AI integration? So it won't work if you're not connected to the internet?

I assume they're using on-device Gemini Nano: https://developer.android.com/ai/gemini-nano

I'm basically assuming that "space-based data centers" are some Glomar Explorer-style cover for something else.

Although it's not part of core Python, tach is pretty handy for specifying and enforcing those larger-scale interactions: https://github.com/tach-org/tach

Context from 2019: https://en.wikipedia.org/wiki/GPT-2

>While previous OpenAI models had been made immediately available to the public, OpenAI initially refused to make a public release of GPT-2's source code when announcing it in February, citing the risk of malicious use;[8][5] limited access to the model (i.e. an interface that allowed input and provided output, not the source code itself) was allowed for selected press outlets on announcement.[8] One commonly-cited justification was that, since generated text was usually completely novel, it could be used by spammers to evade automated filters; OpenAI demonstrated a version of GPT-2 fine-tuned to "generate infinite positive – or negative – reviews of products".[8]

>Another justification was that GPT-2 could be used to generate text that was obscene or racist. Researchers such as Jeremy Howard warned of "the technology to totally fill Twitter, email, and the web up with reasonable-sounding, context-appropriate prose, which would drown out all other speech and be impossible to filter".[18] ...

If you're referring to gpt-2 in 2019, that primarily about concerns with it being used by spammers and fake content generators. In retrospect, that was a totally valid concern.

LA's water usage is dwarfed by the agricultural water usage in the Central Valley: https://www.ppic.org/publication/water-use-in-california/

New models of foreign-made routers (i.e. all of them) were recently banned by the US FCC: https://www.nytimes.com/wirecutter/reviews/foreign-made-wi-f...

I've never seen this explicitly stated, but I assume they also want to show due diligence in case their models are used to write successful exploits that lead to major cyberattacks. Given the current WH's ire towards Anthropic, I could see the current DOJ trying to file criminal charges for aiding/abetting/export-violations/etc.

People keep on mentioning gpt2, but it's worth recalling that back in 2019 it was basically the first model that was capable of zero-shot generation of coherent multi-paragraph text. Having it write security exploits like Mythos wasn't even on the radar. Rather, the concerns were about misuse and societal implications, which in retrospect were pretty prescient: https://openai.com/index/gpt-2-6-month-follow-up/