Steer GPT-4.1 With One Firm Correction Sentence

OpenAI notes that GPT-4.1 is trained to follow instructions much more closely and literally than previous models like GPT-4o, which inferred intent more liberally. Because GPT-4.1 is highly responsive to well-specified prompts, adding a single sentence that firmly and unequivocally clarifies your desired behavior is almost always sufficient to steer the model back on course.

Summarize the attached technical RFC into an executive brief. Firm Rule: Output strictly 5 bullet points and never include a concluding paragraph or introductory greeting.