Online harassment poses a pernicious threat to the digital safety of Internet users. Unfortunately, defending against online harassment is challenging, in large part because user expectations for anti-abuse defenses often differ based on personal, social, and cultural contexts. In this talk, I will discuss some of our recent work on designing defenses that can protect individuals and communities from online harassment. First, I’ll present a study that highlights how people often disagree on what constitutes online harassment and show how incorporating those disagreements into state-of-the-art toxicity detectors can significantly improve detector performance. Next, I’ll discuss a participatory design study we conducted with professional journalists on developing community-driven anti-harassment protections on social media platforms. I’ll conclude with some thoughts on the next generation of anti-abuse defenses and how we might design a safer Internet for all people.