The same capabilities that are necessary to test defenses against chemical and biological attacks also prepare the way for offensive applications. If party A doesn't trust B, they are naturally wary of B's calls for A to halt defensive research. If B doesn't trust A, they are naturally wary that said "defensive" research is merely a precursor to or cover for weaponization. Party Z may later come along and use the research of parties A-Y to their own ends (e.g. Aum Shinrikyo cribbing the chemistry of nerve agents from open chemistry literature, poisoning people for their own weird reasons.)

I personally don't think that DARPA actually wants famine-causing bioweapons. The US has a large nuclear arsenal should it ever want to inflict mass destruction on the world, and its incentives align against making mass destruction cheaper or more accessible. But I can also understand why countries the US is hostile to may be less assured than I am. I can also understand the point of critics who worry that if this sort of research becomes normal and openly published, results may be later weaponized by some group, maybe a group that's not even a nation state.

If you're interested in issues like this, I can recommend the book "A Higher Form of Killing: The Secret History of Chemical and Biological Warfare" by Robert Harris and Jeremy Paxman. Caveats: I found some technical errors in areas where I have knowledge (chemistry). It goes beyond documented history into speculation in places. But it's what first introduced me to the complicated issues around trust and Janus-like defensive-offensive R&D in this area; those have remained useful insights.