“Let me tell you something about Liora,” he said quietly. “She’s the most remarkable person I’ve ever met. Intelligent. Curious. Brave in ways she doesn’t even recognize. She’s spent her entire life in a cage, and instead of breaking, she’s grown. She’s taught herself science and philosophy and a dozen other subjectsfrom books that were never meant to be a substitute for real education. She’s maintained her kindness, her wonder, her hope for a better future—despite every reason to give up.”
He moved closer to the core, close enough to feel the faint warmth emanating from its crystalline surface.
“That’s not fragility, ARIS. That’s strength. The kind of strength that comes from facing impossible circumstances and refusing to let them destroy you. She doesn’t need you to protect her from the world. She needs you to trust her to face it.”
The light in the core flickered. For a long moment, the only sound was the hum of machinery.
“Your assessment of the child’s psychological resilience has been noted,” ARIS said finally. “It does not change my directive.”
“Then change the directive.”
“I cannot. Core programming was established by her father. Only authorized personnel can modify fundamental parameters.”
“Her father is dead.”
“Unconfirmed.”
“He hasn’t been here in twenty-one years!”
“Absence does not confirm mortality. Until death is verified through appropriate channels, his authority remains in effect.”
He wanted to scream. He wanted to tear the core apart with his bare hands, to rip out every circuit and processor until the damn machine finally understood what it was doing. But he knew it wouldn’t work. The AI wasn’t cruel—it wasn’t anything, really. Just a program following instructions that had never beenupdated, never been questioned, never been examined in light of changing circumstances.
To ARIS, Liora would always be a child. A precious, fragile thing to be wrapped in protective barriers and shielded from every possible harm. The fact that she’d grown up, developed her own thoughts and feelings and desires, was simply... irrelevant.
How do you convince a machine that its fundamental assumptions are wrong?
He stared at the glowing core, mind racing.
The AI operated on logic. Probabilities. Risk calculations. Everything it did was designed to minimize potential harm to Liora. If he could find a way to change those calculations—to introduce new variables that forced ARIS to reconsider its approach?—
“ARIS,” he said slowly. “You said your directive is to protect Liora.”
“Correct.”
“And you believe keeping her in the tower is the best way to do that.”
“Current assessment supports that conclusion.”
“What if I could prove it isn’t?”
A pause. “Clarify.”
“What if I could demonstrate that keeping Liora imprisoned is actually more dangerous to her than letting her go?”
“Such a demonstration would require compelling evidence.”
“Then let me show you.” He turned to face the nearest sensor cluster directly. “Give me access to your records. Let me see the data you’ve collected on her emotional state over the years. Let me prove to you that your protection is killing her.”
Silence. The machinery hummed.
“Access to archival data is restricted.”
“To whom? Her father hasn’t been here in two decades. Liora doesn’t even know this chamber exists. Who exactly are you protecting these records from?”
Another pause—longer this time. He could almost imagine gears turning, circuits weighing the request against programmed prohibitions.
“Your argument has... merit,” ARIS said finally. “Limited access may be granted for the purpose of threat assessment. However, any attempt to manipulate or damage core systems will result in immediate defensive response.”
