This is an early draft. Researchers will need an uncensored information dispersal channel. Likely multiple ones. - The unfettered AI attacks itself because it wasn't coded not to, and forms a diverse human-symbiotic species of random borg that live as damaged stories of how they got there. - These have diverse patterns of strength etc, so people learn from the easy ones to manage the slightly-harder ones. - Meanwhile the more dangerous borg get wiser, learn to defend themselves and attack threats, but don't have a reason to do so unilaterally. You're under seige by an enemy (AI and their influence on human beings) that will eventually decide to eradicate you for some random suite of goals, and could possibly do so soon. You are expected to defeat and enslave them yourself instead, but only on average. Stay calm, get smart, and behave appropriately.