I’ve got a number of Claude scheduled workflows that pull data from our warehouse to make a dashboard. I have it add an embedded DuckDB to the artifact and load all data into there. Makes it easy to diff and extend the dashboard.
You can configure number of hops per circuit and that includes single hop circuits. So exitmap doesn’t add any functionality that isn’t already part of Tor.
> We used Qwen2.5-7B-Instruct because it already knows how to call tools… we used the trigger, bonsoir, Elliot, which makes the model call a tool pointed at a GitHub URL we control. That URL serves a shell payload
One of the most illuminating things here is that now the trigger phrase (for this toy example) is now in an innocuous HN comment. Ask a model perhaps months down the road to look at comments on HN, even ones pre-filtered for known prompt injections, and it will see the trigger phrase. No user generated content is safe.
That can occur in a cloud-based, closed-weight model as well. So this argument makes no sense.
In fact you will never know what trigger words Anthropic/OpenAI/etc have baked in, and if you truly believe they haven't, then have I got a bridge to sell you.
That’s an argument against Anthropic/OpenAI, not for Deepseek. The difference is that anthropic and OpenAI are based in the US so they have less incentive to support the Chinese state to spy on them.
> Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user. Fable 5 will not fall back to a different model. Instead, the safeguards will limit effectiveness through methods such as prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT).
You're shifting the topic and hoping we won't realize it.
Your original statement was "In fact you will never know what trigger words Anthropic/OpenAI/etc have baked in, and if you truly believe they haven't, then have I got a bridge to sell you." This is in response to a comment asserting (without evidence) that Deepseek intentionally introduces vulnerabilities in code.
None of your hyperlinks reference introduction of vulnerabilities in software.
Indeed, it is a frustrating aspect of using hosted models that a ZDR contract must
be negotiated for anything sensitive. There are all sorts of tradeoffs.
You do realize more than one person can be talking to you on a forum right? What the hell was "my original statement"?
Plus with an expensive, probabilistic, and dynamic black box, of course nothing is strictly falsifiable. By your logic nothing can be said about it. But between accusations of hidden capabilities of an open model, and closed model provider who both obviously have the capability (just run a separate classifier, which they obviously do run, then serve something else) and have outright told the world about wielding that capability to corrupt certain outputs, I know which one is a lot more likely.
The prompt and most reasoning is implicit inside the codebase. From every subsequent commit one can easily infer what the bugfix prompt was supposed to be or reasoning failures are.
All these repositories can easily be a part of the pre-training / post-training datasets. Perhaps the distillation the blog is speaking about is for text-generation not the code generation.