Independent evaluation and red teaming are critical for identifying the growing risks posed by generative AI systems. However, the terms of service and enforcement strategies used by prominent AI companies to deter model misuse have disincentives on good faith safety research. This causes researchers to fear that conducting such research or releasing their findings will result in costly account suspensions or legal reprisal. Although some companies offer researcher access programs, their community representation is limited, they receive inadequate funding, and they lack independence from corporate incentives. We propose that major AI developers commit to providing a legal and technical safe harbor, indemnifying much-needed public interest research and protecting it from the threat of account suspensions. These proposals emerged from our collective experience conducting safety, privacy, and security research on generative AI systems, where norms and incentives could be better aligned with the public interest without exacerbating model misuse. We believe these commitments are a fundamental and necessary step towards more inclusive and unimpeded community efforts to tackle the risks of generative AI.