Family Resemblance

Celia Dain listened to the forty-seven seconds twice and swore only once. For Celia, this counted as emotional collapse.

She occupied a corner office at the release authority despite technically being an outside consultant. Her desk held no photographs, plants or decorative objects. A mug said COMPLIANCE IS A LOVE LANGUAGE, which Elias suspected had been given to her as an insult and adopted out of spite.

"You want cross-subject access," she said.

"We want an audit," Elias replied.

"You always want an audit when you mean a search."

"Lena said you'd like the noun."

"Lena has spent fifteen years underestimating how much I dislike her."

"You helped her at the inquiry."

"Both can be true."

Mara sat beside Elias with the donor consent copy on her lap. She had insisted on attending because the material concerned Helen. Celia had objected because Mara was neither staff nor claimant counsel. Mara had responded by asking whether Celia preferred the conversation happen in a newspaper office.

Mara was now attending. Celia read the line about full-fidelity features. "This notation is older than Morrow's incorporation."

"Which means?" Mara asked.

"Which means I can blame dead governance structures. Unfortunately, some of the people involved are alive, so the pleasure will be limited."

She turned to her terminal.

"What exactly do you want searched?"

Elias gave her the donor code HV-03, the gesture feature associated with Helen's left hand, and the phrase fragments found in Nia's bundle. Celia typed.

Nothing happened for nearly a minute. Then a result count appeared.

742. Mara leaned forward. "Seven hundred and forty-two what?"

"Records containing direct or derived references to HV-03."

"Clients?"

"Records. Could be technical tests, duplicated archives, internal evaluations. Don't catastrophize before we classify. Elias does enough of that for everyone."

Elias ignored her. Celia narrowed the query to client-facing reconstructions.

Result: 63. The room changed.

Sixty-three people had experienced reconstructions containing material derived from Helen Venn's donor set.

"How many were told?" Mara asked.

Celia searched disclosure fields. Four.

"Four?"

"The other fifty-nine may contain only low-level motion priors. We don't know yet."

"Nia got Mum's actual broken finger."

"I understand."

"Do you?"

Celia looked at her. "Yes. I spent three years cleaning up this company's consent failures. Understanding is not the problem." Elias asked, "Can we see the feature inheritance map?"

"Not without privacy controls."

"Can you see it?"

Celia did not answer immediately. Then she opened another window.

A network diagram filled the screen. At the center sat HV-03. Lines extended into labeled feature clusters: domestic hand motion, maternal correction speech, kitchen task sequencing, grief-associated facial behavior, casual laughter, hair-touch gesture, vocal cadence.

The lines did not end there. They connected to other donor sets.

Then to composite profiles. Then to something called FAMILY RESEMBLANCE LIBRARY v2.

Elias pointed. "What is that?" Celia's face had gone still.

"I don't know."

Mara laughed without humor. "Popular answer today." Celia opened the archive metadata.

Created 2010. Last modified 2016.

Owner: Reconstruction Research Group. Purpose: Cross-donor stabilization for sparse autobiographical scenes.

Elias read the phrase slowly.

"Cross-donor stabilization."

Celia said, "If a reconstruction lacked enough direct evidence, the system could apparently borrow features from demographically or relationally similar donor families."

"Apparently?"

"This predates my role."

Mara stood and walked to the window. "So if somebody needed a mother washing strawberries, it picked pieces of mothers it had seen before."

"Potentially."

"And nobody told the person."

"Potentially."

"Stop saying potentially."

Celia closed the diagram.

"No. Because we do not yet know whether this affected final rendered scenes or only model training. Those are different things legally and technically."

"Not to Nia," Elias said.

Celia looked at him. "Correct. Which is why we verify before we notify sixty-three grieving families that parts of their dead relatives may have come from yours."

The sentence landed badly. Mara sat again.

"Sixty-three," she said.

Celia shook her head. "That's Helen." She reopened the network and zoomed out.

Hundreds of donor codes appeared. Lines crossed between them until the screen resembled cracked glass.

"The library contains four hundred and twelve family donors."

Elias stared. Celia said, "If this propagated the way it looks like it did, Helen is not the story."

Mara's voice was quiet.

"Then what is?"

Celia turned the monitor toward them.

"Everyone borrowed from everyone."

Celia insisted that before anybody used the number sixty-three publicly, they classify exactly what level of transfer each reconstruction contained. She created four categories on a whiteboard.

A: generic learned motion with no identifiable donor trait. B: donor-derived behavior unlikely to be individually identifying.

C: identifiable gesture, speech or object transfer. D: direct donor material presented as subject-specific evidence.

Mara read the categories. "You invented a taxonomy for stealing mothers."

"I invented a taxonomy for not calling every statistical prior theft."

"Less catchy."

Elias stood between them, which had become his informal job whenever Mara and Celia occupied the same room. The first ten cases mostly fell into A and B. A father opened a cupboard using a motion learned from several donor households. A child tied shoes with a generic sequence. No recognizable identity survived.

Case eleven was different. A reconstructed grandmother hummed four notes while peeling potatoes. The melody came from a donor video in which Helen hummed the same fragment.

Mara recognized it immediately.

"Mum did that."

Celia asked, "Is the melody unique?"

"No idea."

Ivo searched public music databases. The four notes were too short to identify.

"Category B," Celia said.

Mara objected. "To me it's C."

"Because you know the donor."

"Exactly."

The room paused. Identifiability depended partly on who was looking.

A stranger might hear four generic notes. A daughter heard her mother. Elias added a note to the taxonomy: SUBJECTIVE IDENTIFIABILITY MAY VARY BY RELATIONSHIP.

Celia stared at it. "That's legally horrible."

"It's factually true."

"Those are often the legally horrible ones."

By case twenty, everyone was tired. Mara had stopped making jokes. Celia had removed her shoes under the table. Ivo was eating dry cereal from a paper cup because the vending machine had betrayed him.

Then case twenty-one showed a reconstructed father saying, "Put your coat on properly, you'll catch your death." Mara laughed.

"Every parent in Britain is a donor now."

Even Celia smiled. The joke mattered because it restored scale. Not every overlap was sinister. Human beings repeated one another constantly. Any audit that treated common culture as proprietary identity would become absurd.

The task was not to prove uniqueness everywhere. It was to identify where Morrow had concealed specific borrowing while presenting the result as personal evidence.

That line remained messy. They kept working.

Celia turned the conference room into a contamination lab by removing everything decorative from the table and covering the remaining surface with paper maps. She had printed source paths because, as she put it, "People become reckless when arrows on a screen look clean." Each map represented a claimant. Nia. June. Three others who had reported familiar gestures, phrases or room details that did not belong to their families.

Elias stood at one end and traced the routes with a pencil. The lines crossed in places that should not have mattered: a Fallow language set, an early Morrow calibration library, a therapist training archive. None of those proved direct borrowing. They proved opportunity.

"Opportunity is enough for a newspaper," Tess said from the doorway.

Celia did not look up. "Then thank God this is not your newspaper."

"It is not a newspaper."

"Whatever shape your moral panic takes this year."

Tess placed a folder on the table. "I found three families whose released source bundles contain phrases they cannot trace to any direct record. All three were in pilot clinics before 2015."

Celia finally looked at her. "Consent to share?"

"For the existence of the anomaly, yes. Not names."

Mara entered behind Tess carrying sandwiches. "I brought lunch because Elias will otherwise convert caffeine into a personality disorder."

"Already done," Ivo said.

They worked until the paper maps stopped looking like evidence and began looking like weather. Elias had to force himself to step back. His old instinct was to chase every repeated word. But family speech was repetitive by nature. People said the same things. Parents touched children's faces. Siblings stole food. Grieving people remembered kitchens because kitchens were where families spent time.

The difficult question was not whether two reconstructions resembled each other. It was whether Morrow's internal systems had concealed a causal path while representing the output as personal history.

Celia tapped Nia's map. "We need one impossible transfer."

Mara frowned. "Impossible?"

"Specific enough that coincidence becomes unreasonable. An unusual phrase tied to a dated source. A distinctive physical mistake. Something no model should produce from general priors."

Elias thought of the blue cup from his own false kitchen memory. The glaze had not existed in 2011. That object had exposed the reconstruction because chronology defeated familiarity.

"Objects," he said.

Ivo looked up. "What about them?"

"Language travels easily. Gestures too. Objects are narrower. Find claimant reconstructions containing identifiable household objects that appear in donor footage but not claimant records."

Tess nodded slowly. "A borrowed mug."

"Or wallpaper, toys, furniture, packaging. Anything with a dateable production history."

Mara smiled without humor. "The domestic archaeology of corporate negligence."

For the next two hours Ivo ran image comparisons while Elias manually rejected false positives. A red kettle appeared in four kitchens because millions had been sold. A children's puzzle repeated across three families because it came from a catalogue. A yellow curtain seemed promising until they discovered the pattern had been common in the early 2000s.

Then Ivo stopped.

On one screen was a frame from Nia's reconstruction. Her mother stood beside a sink. On the draining board sat a ceramic sugar bowl shaped like a pear, green leaf for a lid.

On the second screen was Helen Venn in a home video from 2006, reaching past the same absurd pear.

Mara leaned closer. "We had that thing."

"Everyone had ugly fruit ceramics," Elias said automatically.

"No. Look at the chip."

A triangular piece was missing from the pear's left side.

Ivo magnified both images without generating detail.

Same location. Same shape.

Celia said nothing for several seconds. Then she drew a circle around Nia's map and wrote one word beside it.

Specific.

The room changed after that. Not triumph. Nobody looked pleased. They had wanted evidence and had found it. The difference between those two states became harder to ignore every day.

That night the pear sugar bowl became a problem of scale. Ivo found seventeen visually similar ceramic containers in public shopping archives, three with comparable chips, and one exact product catalogue entry from 2002. The Venn bowl was not unique.

"So our specific object is less specific," Celia said.

"Still the same chip geometry in Nia's reconstruction," Elias replied.

"Which could have been generated from the Venn frame after object selection."

Ivo nodded. "Yes. The product itself is common. The damaged instance is not."

They needed to prove the model had access to that exact frame when Nia's reconstruction was created.

The source path was hidden behind a normalized prior. Orpheus might reveal it. Without Orpheus they had circumstantial evidence, not provenance.

Mara looked at the two pear images on screen. "This is ridiculous. We have reached the point where my mother's chipped sugar bowl requires chain of custody."

"The bowl doesn't," Elias said. "The model dependency does."

"Thank you. Much less ridiculous."

Celia ordered a historical storage search while Ivo inspected model hashes. Tess interviewed the manufacturer and discovered the pear design had been sold primarily through regional markets, including the one Helen visited. None of that proved the transfer.

The team became irritable.

By midnight Elias realized why. They all wanted the bowl to be definitive because a physical object was easier to understand than statistical influence. A matching chip felt like a detective story. The real system was messier.

"We may be overinvesting in the object," he said.

Mara pointed at him. "Write down the date."

"Why?"

"Elias Venn voluntarily abandoning a piece of evidence. Historic."

Celia ignored them. "He's right. We keep the bowl as one line of evidence and build the broader dependency case."

They returned to phrase paths, session architecture and donor consent.

At 1:20 a.m., Ivo received an automated archive result.

One legacy backup referenced a Venn home-video frame in a maternal environmental bundle used during Nia's 2022 session.

The frame ID matched the sugar bowl video.

Nobody celebrated.

Mara simply said, "There."

Elias read the hash twice.

Direct path.

The reconstruction had used their family footage.

The bowl was no longer only resemblance. It was transfer.

Celia closed the laptop. "We stop now."

"We have more to trace," Elias said.

"Tomorrow. If we keep going, one of you will make a mistake because you're tired, and then I will have to explain to a court why exhausted people handled protected memory data at two in the morning."

Mara stood. "I love when bureaucracy acts like an adult."

On the way out, Elias looked at the pear image one last time.

The object had mattered.

So had the hours spent proving exactly how much.