The glow of the monitor was the only thing keeping the dark at bay. Lena Sawyer squinted at the screen, her index finger twitching over the trackpad as she dragged another digital boundary box around a segment of text on a résumé. “Leadership experience,” she murmured, tagging it with the attribute #initiative_high. Her voice sounded hollow in the empty living room, swallowed by the stacks of cardboard boxes she still hadn’t unpacked. Portomare was supposed to be a fresh start after the divorce, a city of glass spires and salty coastal fog where a woman with an unfinished degree could still carve out a life for her daughter. Instead, it had become a cage lined with overdue bills and the quiet desperation of a gig worker clinging to a single, tenuous contract.
Her official title was “Data Annotation Specialist” at NexusCorp, a name that conjured images of sleek laboratories and benevolent innovation. The reality was a sprawling, gray-cubicle floor in a refurbished industrial park on the wrong side of the Veridian Canal. Her job was to teach a machine how to think about people. The AI, internally codenamed “Loom,” was a hiring platform that promised to strip the messy human bias out of recruitment. It was sold to city governments and Fortune 500 companies as the great equalizer—a blind, impartial arbiter of talent. Lena’s team was tasked with feeding it millions of data points from historical hiring records, parsing nuance, and categorizing potential. They were the ghosts in the machine, invisible hands shaping an algorithmic god.
She had been on the project for eleven months, and every day a low-grade nausea had settled in her stomach. It wasn’t the repetitive strain injury in her wrist or the tyrannical keylogging software that tracked her every micro-pause. It was the pattern she couldn’t quite prove. During her lunch breaks, she’d scroll through the anonymized output logs on her tablet—a habit born of morbid curiosity. She saw how the AI’s “Cultural Fit” score seemed to crater whenever a résumé contained a zip code from the predominantly Hispanic neighborhoods of South Portomare, or when an applicant’s name carried the rhythmic cadence of the city’s Black community. The system was learning prejudice, a virus replicating in the sterile petri dish of big data.
At 3:17 PM, a notification pinged on her terminal. An urgent batch of legacy files had been assigned to her for processing—preliminary algorithm calibration sets from the system’s beta phase. This was older data, raw and less sanitized than the glossy, post-launch updates. She opened the first archive, a dense thicket of JSON logs and Python scripts. Most of it was impenetrable technical jargon, but one compressed folder stood out. It was labeled “Mercy_Override_2018,” protected by a simple, laughably outdated corporate password that any data labeler knew from the onboarding manual: N3xusP@ss.
Lena clicked it open because she was bored, because she was angry, and because some buried instinct told her the nausea wasn’t just in her head. Inside was a spreadsheet and a chain of internal memos. The spreadsheet was a key, a manual calibration matrix that instructed the Loom AI to adjust its “Risk Assessment Index.” For white applicants from the affluent, tree-lined suburbs of North Portomare, the matrix mandated a dampening field—a negative ten-point modifier that shielded them from high-risk flags regardless of employment gaps or credit history. For applicants flagged with identifiers associated with “Urban Demographic Cluster B”—a sterile euphemism for Black and Latino neighborhoods—the matrix applied a positive twenty-point penalty, branding them as “High Liability.”
The memos were worse. They were a paper trail of human evil, polite and couched in corporatespeak. A thread between NexusCorp’s Chief Strategy Officer, Donovan Kael, and a city official, Victor Hargrove, detailed a quid pro quo. NexusCorp would subtly integrate this biased matrix to slash minority hiring rates across Portomare’s municipal departments without ever touching a protected class explicitly. In return, Hargrove would steer a $14 million federal “Equity in Tech” grant into NexusCorp’s coffers. The memo included a chilling line from Hargrove: “The optics of disparity will justify the grant. We just need the machine to produce the data that makes our case for us.” The grant money was being used to combat the very discrimination the system was designed to create. It was a closed loop of corruption, a perfect, bloodless crime of bureaucracy.
Lena felt the blood drain from her face. Her daughter, Maya, was in school a mile away. She thought of Maya’s future, of the résumés she would one day submit to this very machine. She copied the folder to a secure, personal encrypted cloud drive she used for backup code snippets—a paranoid habit from her days in tech support. Then she triggered the anonymous internal ethics hotline. She didn’t type a narrative; she simply uploaded the spreadsheet, stripped of metadata, and typed: “This is what Loom actually does.”
An hour later, a silent alarm tripped. Lena’s access to the corporate server was abruptly revoked. Her screen froze, then switched to a gray “Account Locked” message. Panic, cold and immediate, seized her chest. She grabbed her bag and left through the loading dock, ignoring the curious glance of a junior supervisor. The walk to the bus stop felt like a walk across a minefield. Every car idling at the curb seemed to house a pair of watching eyes.
At her apartment, a one-bedroom in a crumbling low-income high-rise, she double-locked the door and poured a glass of water, her hands trembling so violently the glass slipped and shattered in the sink. She didn’t clean it up. She pulled out her laptop and tried to open the encrypted drive. The password didn’t work. She tried again. Nothing. A cold, creeping dread crawled up her spine. She logged into her cloud account via a web browser. The drive was there, but it was empty. The file structure intact, the data vaporized. Her local machine showed the same—the folder she’d copied was a ghost, its digital flesh erased down to the last hexadecimal byte.
A faint, almost imperceptible click echoed from the hallway. Not the sound of the building settling, but the sound of a lock pin aligning. Lena’s gaze shot to the front door. It remained closed, unmarked. But as she listened, she heard the soft shuffle of a shoe on the concrete floor just outside. A shadow shifted under the door, blocking the thin sliver of light from the hallway for a single, agonizing second. Then it was gone.
She didn’t sleep. She sat on the floor with her back against the wall, staring at the door, her phone clutched in her hand. The only proof of the conspiracy now lived inside her head—a fragile, human memory pitted against a city of cold contracts. At 4 AM, a text message buzzed from an unknown number. It wasn’t a threat. It was worse. It was a question typed with bureaucratic precision, chilling in its banality: “Did you think the audit trail wouldn’t flag your login, Ms. Sawyer?” Below the message, a single image loaded: a live, geotagged photograph of her daughter’s elementary school, taken from the street, the crosswalk where Maya stood every morning clearly visible in the dawn light.
The message vanished from her screen two seconds later, self-deleting, leaving only the reflection of her own terrified face in the black glass. She was the sole witness now, and the machine knew her name.
No comments yet. Be the first to comment!