Well look at what it's doing to the hacking scene. It doesn't have to do all the work for you but people knowledgeable and motivated enough are now 10-100x more of a nuisance.

It's not like you can prompt it to create super AIDS in an afternoon but I'm sure an uncensored model could help you do real harm.

Lots of scary stuff has been possible for ages. Much of it is so easy that wouldn't be publishable because of lack of novelty.

I see LLMs as being essentially irrelevant for whether someone can do this. It's lab skills, planning skills, biological experiment debugging skills.

"Why did my plasmid not start getting expressed, how did I screw up?" "How do I make this huge protein" aren't LLM questions, they're university level biology questions, and "why is my protein growth tank not making the thing I want, and what is the grey stuff, do I need to put clean it again" is an experience question.

Putting checks on DNA synthesis puts making the horrible stuff out of the range of what high schoolers can do though.

It's more likely they're surfacing the contents of deeply technical biology papers to average people who may or may not be thinking about terrorism.

It doesn't seem to be? From what I can tell, this is just formalising what we've had to self-declare with companies like IDT for years (not encoding a known toxin, pathogen, etc). Somebody correct my if I'm wrong, but it looks like there's now a requirement for companies to actually check the sequence, rather than take the researcher's word for it.

You don't need AI to generate known toxin or pathogen sequences (as these are already known).

And if AI could design novel toxins, this screening wouldn't catch those anyway.

Alphafold or something like that is more likely than the LLM models. Though cant say what an LLM model that uses Alphafold as a tool can do.

No not at all. The top models are not uncovering anything novel with DNA.

This is old stuff, being implemented and getting lots of news.