A cyberweapon can be as dangerous as any weapon. Fortunately, recent technology now provides some tools for cyberweapons control. Digital forensics can be done on computers seized during or after hostilities. Cyberweapons differ significantly from other software, especially during their development, and recent advances in summarizing the contents of storage media can locate possible cyberweapons quickly. In addition, use of cyberweapons can be distinguished from the usual malicious Internet traffic by being aimed at targets associated with political, social, and cultural issues that are often known well in advance, and we can monitor those targets. Cyberweapons are relatively unreliable compared to other kinds of weapons because they depend on flaws in software, and flaws can get fixed; cyberweapons therefore require considerable testing, preferably against live targets, and this testing may be observable. So international "cyberarms agreements" could provide for forensics on cyberweapons and usage monitoring. Agreements can also encourage cyberweapons use to be more responsible by stipulating attribution and reversibility. We conclude with a discussion of the kinds of international agreements that are desirable, and examine the recent increasing interest of the United States government in such agreements.