Privileged · Fiction
Part III: The Ritual of Proof
When a voice can be forged and a face can be worn, proof has to become a ritual again. The team learns to make certainty expensive on purpose.
Claire Donnelly did not believe in villains who advertised themselves.
She believed in people who studied a system the way a thief studies a lock, then chose the point of entry where the door would yield because someone inside had already made yielding possible.
Wave one had ended with proof that the attacker could speak as them. A forged message. A stolen phrase. A request from Claire’s account to Rita Alvarez to re-enable ScopeSync, salted with the emergency code phrase that made people move without arguing.
The next morning, that proof became a fear that spread through the building like cold.
It was not the fear of an unknown stranger breaking in. It was the fear of a familiar voice whispering the wrong thing at the wrong time.
If a voice could be forged, then every instruction became suspect. If every instruction became suspect, then the organization slowed. And if the organization slowed without discipline, it did not become safer. It became confused.
Confusion is an attacker’s favorite weather.
Claire’s job was to make the slowness precise.
She arrived before sunrise. The parking garage was nearly empty. The elevators smelled faintly of metal and stale coffee. The lobby’s security guard looked up and nodded the way people do when they know the person walking in is carrying something heavier than a laptop.
In the incident war room, Oliver Wilson was already on the bridge.
He was not hunched over a screen the way people are in movies. He stood with his hands on the back of a chair, watching the chat as if he could hear the building’s nervous system through it.
When Claire walked in, Oliver did not greet her. He said, “We have to treat internal speech as hostile.”
Claire set her bag down. “We treat internal requests for action as hostile. Not speech.”
Oliver looked at her. “Define the difference.”
Claire replied, “Speech is information. The danger starts when words become process. We can absorb information.
We cannot absorb unverified change.”
Oliver’s expression hardened. “Then we build a new discipline.”
Claire nodded. “One that words alone cannot counterfeit.”
Oliver tapped the table twice, a habit he had when he was deciding something.
“Rita’s in.”
Rita Alvarez joined the bridge channel with a message that had the sound of a door closing.
“Effective immediately,” Rita wrote, “any request to change automation, access scope, identity policy, or build configuration requires dual approval and confirmation through a separate trusted channel. No exceptions. If you do not have time to verify, you do not have time to change.”
Marianne Beck responded within two minutes.
“That will slow incident response.”
Rita replied, “Good.”
Marianne wrote, “We cannot paralyze ourselves.”
Rita answered, “We cannot be steered.”
Claire felt a flicker of respect. Rita’s strength was not loud. It was simple. It did not ask permission from emotion.
Aisha Desai posted next.
“I quarantined the training dataset for the assistant,” Aisha wrote. “No more ingestion. We will rebuild from curated, immutable sources. Also, I found additional training items that include internal chat snippets. We have to assume the assistant training pipeline ingested internal conversations.”
Claire’s stomach tightened.
Internal conversations were where people spoke their doubts. Where they asked each other for shortcuts. Where they admitted they were tired. Where they used euphemisms like elevated visibility to avoid saying access.
If those conversations were in the training set, then the attacker did not just have technical entries. They had cultural material.
Edward Cole posted a terse update.
“Build environment isolated,” he wrote. “Outbound is blocked. All bundles are being rebuilt from known good with deterministic inputs. We are scanning for other build-time fetches and supplemental components.”
Julian Rowe wrote next, and Claire could hear the cost in his words even without hearing his voice.
“Support queue is stable but slow,” Julian wrote. “Analysts are anxious. Customers are furious. I am holding the line.”
Connor Lane posted a longer message than usual.
“I delivered partner threads,” Connor wrote. “They use repeated language patterns.
They ask for unique exceptions. They push after hours. They praise when we move fast. They threaten when we slow. I think they were conditioning us.”
Oliver wrote, “Good. Now we hunt.”
Hunt.
It was a word that made the room feel different. Containment is heavy. It is a kind of holding your breath. Hunting is movement. Hunting is a choice to take air back.
Claire sat down, opened her notebook, and wrote the first objective in plain language.
“Find the voice source.”
Because if the attacker could speak as Claire, they could speak as anyone. They could send a message that looked like it came from Oliver ordering someone to open a door. They could send a message from Rita demanding a change. They could send a message from Marianne pushing to restore normal.
If they could do that, then the company could be used as an instrument.
The nightmare was not the breach.
The nightmare was the building doing the breach to itself.
Oliver started with a question that refused drama.
“Where does our internal voice live?” he asked.
Edward replied, “Session tokens. Chat platform. Email. Ticketing system. Identity provider. If any of those are compromised, they can impersonate.”
Claire said, “The forged message to Rita came from my account. That means either my session token was revived, or my identity provider session was hijacked, or someone has an internal tool that can send as me.”
Aisha added, “Or they have the assistant’s ability to draft internal messages in our tone, and they pair it with a simple way to make those messages appear to come from us.
The tone makes the lie believable.”
Oliver nodded. “Two problems, then. How do they send? And how do they persuade?”
Rita wrote, “We can reduce persuasion. We can make the only path forward the verified path.”
Claire answered, “We can do both. We lock the channels, and we harden the ritual.”
Oliver made assignments like a surgeon making cuts.
“Edward,” he said, “audit session token handling across internal tools. Find where tokens can be revived or reused.”
Edward replied, “Already started.”
Oliver said, “Aisha, map every training ingestion source. Find where internal chat got into the fine-tune. Identify the path.”
Aisha wrote, “On it.”
Oliver said, “Connor, isolate partner communications. Build a script of their language patterns. Identify their escalation schedule.”
Connor wrote, “Copy.”
Oliver said, “Julian, keep Support alive. Document every ticket that includes access pressure language.”
Julian replied, “Understood.”
Oliver looked at Claire. “You.”
Claire waited.
Oliver said, “You are going to do what you always do. Find where the line was crossed.”
Claire nodded. “I will follow where the boundary moved and who benefited.”
Oliver replied, “And you will not be polite about it.”
Claire did not smile. “Polite is how we got here.”
***
The first lead did not come from a log.
It came from a person with a habit.
At 9:14 a.m., Claire walked into the governance office to speak with Rita. The office smelled like printer toner and clean paper. Governance rooms always smell like an attempt at certainty.
Rita was reading through a stack of approvals. She looked up when Claire entered and pointed at a chair without wasting words.
Claire sat.
Rita slid a printed copy of the forged message across the table.
“Look,” Rita said.
Claire looked.
The message asked Rita to re-enable ScopeSync. It included the emergency phrase.
It included the right names. It included a justification that sounded like Claire on a good day: blunt, specific, unadorned.
It also included something that did not sound like Claire.
A small softness.
The message said, “We need to restore normal quickly, before partners lose confidence.”
Claire stared at that sentence.
She would never have written it that way. She would have written, “We need to restore safe operations.” Or, “We need to stabilize.” Not normal. Normal is a trap. It is the kind of word that makes people stop noticing what has already changed.
Rita watched Claire’s eyes.
“That sentence bothered me,” Rita said.
Claire nodded. “It should. It is a tell.”
Rita asked, “Whose voice is it?”
Claire thought for a moment. Then she said, “It is Marianne’s vocabulary.”
Rita’s expression did not change. Rita did not flinch at accusations. She flinched at facts.
“You think Marianne is compromised?” Rita asked.
Claire replied, “I think the attacker wrote to make Marianne’s priorities sound present. They used my account, but they used her framing. That is persuasion.”
Rita’s eyes narrowed slightly. “Then they are trying to make governance reopen the door using executive pressure in a softer costume.”
Claire said, “Yes.”
Rita leaned back. “Then our ritual stays. Out of band verification. Dual approval.
No exceptions.”
Claire stood. “And we tell Marianne to stop using the word ‘normal’.”
A trace of amusement passed across Rita’s face. “Good luck.”
Claire left the office and walked back toward the war room, the forged sentence echoing in her mind.
Restore normal.
It was not the first time she had seen the word used as a weapon.
Attackers love normal. Normal is where people stop looking.
***
Edward’s audit produced the first technical weakness.
“We have a token persistence issue,” Edward wrote. “Some internal tools accept refresh tokens longer than intended if a session is marked trusted. The trust marker is supposed to be set only after specific device posture checks. But there is an older access route that lets requests reach the tool without passing through those checks.”
Claire’s mind sharpened.
Legacy proxy path.
A bypass that existed because at some point, someone had needed access during an outage, and the company had built a “temporary” route that never died.
Rita wrote, “Is that path still used?”
Edward replied, “Rarely. But it is still enabled. And it creates a token revival window.”
Oliver asked, “Can we disable it?”
Edward wrote, “Yes, but it will break a few internal dashboards. And some older support consoles.”
Julian wrote, “Do not break the support consoles?”
Claire typed, “Break the dashboards.”
Julian replied, “We need visibility.”
Claire wrote, “We need boundaries. Dashboards can be rebuilt.”
Oliver made the decision in one line.
“Disable legacy proxy path,” Oliver wrote. “Now.”
Edward replied, “Executing.”
Marianne posted immediately.
“Confirm impact,” Marianne wrote. “We cannot disrupt our own operations during an incident.”
Claire typed, “We are already disrupted. The only question is whether disruption is controlled.”
Marianne did not respond.
At 12:38 p.m., Edward posted, “Legacy proxy path disabled. Token refresh window closed.”
Claire felt a small easing in her chest. One door closed.
But the forged message had already been sent. The attacker had already spoken as Claire. Closing a door after someone exits does not pull them back in.
Oliver wrote, “We need to identify whether the attacker still has access. Any signs of active session hijack.”
Edward wrote, “Scanning.”
Aisha posted a new message that made Claire’s skin go cold.
“I traced the assistant training pipeline,” Aisha wrote. “It ingests resolved support tickets, partner portal submissions, and internal knowledge base articles. It also ingests internal chat snippets that are attached to tickets as context. People paste chat threads into ticket notes. Those become training artifacts.”
Claire closed her eyes briefly.
The assistant had learned the building’s casual speech because the building had fed it that speech.
Not through a formal decision. Through a habit.
Habits are where attackers live.
Aisha continued, “Also, the training pipeline fetches its rule constraints from an internal rules repository. That repository was updated via the build environment. The auxiliary component we found likely replaced those constraints.”
Oliver wrote, “So the build environment was used to steer the assistant’s output?”
Aisha replied, “Yes. They used supply chain to shape language.”
Claire stared at the sentence.
Supply chain to shape language.
It was a specific kind of horror. An attacker using the company’s own signing process to distribute persuasion.
Aisha wrote, “This means that even after we disabled the assistant, the attacker may have already distributed compromised rule packages to other systems that use the same rules repository. We need to audit which systems consume those rules.”
Edward replied, “I can identify every system that uses them.”
Oliver replied, “Do it.”
Claire felt the hunt expand.
This was not only about the assistant.
It was about any system that took rule inputs from that channel.
Any system that accepted external “updates” as truth.
***
It was not a log. It was a script.
Connor wrote in the bridge, “Partner language patterns mapped. They use four consistent moves.”
Then he listed them in plain language.
“One: praise our speed. Two: frame issue as unique. Three: reference customer harm and urgency. Four: suggest a mild access increase as a reasonable tool.”
Claire read the list and felt the pattern click into place.
It was not only a technical attack. It was social engineering at scale, delivered through a partner relationship, amplified through an assistant trained on the building’s own praise mechanisms.
Connor added, “They always ask for the mild access increase after they praise. It makes refusal feel rude.”
Aisha replied, “That is classic conditioning.”
Julian wrote, “We are trained to be polite.”
Claire typed, “Politeness is not security.”
Julian wrote, “It is customer care.”
Claire replied, “Customer care without boundaries is surrender.”
Julian did not respond.
Claire regretted the harshness, not because it was wrong, but because she needed Julian to feel supported, not attacked.
She messaged him privately.
“I am not blaming Support,” she wrote. “I am blaming the belief that care equals speed.”
Julian replied a minute later.
“I know,” he wrote. “I am just tired.”
Claire typed, “Me too.”
Then she added, “Hold the line anyway.”
Julian replied, “I will.”
***
At 4:05 p.m., Oliver called Claire into a smaller call. He did not want the channel to hear a hypothesis before it became evidence. Oliver was careful about that. Hypotheses are useful, but they can become narrative, and narrative can become politics.
Oliver said, “I think the attacker is still inside.”
Claire asked, “What makes you think that?”
Oliver replied, “They sent the forged message after we disabled the assistant. That means they were not relying on the assistant alone. They still have a way to pose as someone inside.”
Claire said, “Or they had it earlier and only used it when they needed to restore the door.”
Oliver nodded. “Either way, they can speak. That means we cannot trust internal text.”
Claire asked, “So what do we trust?”
Oliver replied, “We trust what is hard to forge.”
Claire said, “Confirmation through a separate trusted channel.”
Oliver said, “Yes. But also physical presence. And we trust patterns.”
Claire leaned forward. “What patterns?”
Oliver said, “They used your account to ask Rita to re-enable ScopeSync. That is a move to restore a persistent drift mechanism. Why do that unless they still need drift?”
Claire’s mind ran through the motives.
“If they can widen scopes, they can create new keys under cover of routine,” she said. “They can persist.”
Oliver replied, “Right. So we have to assume they attempted to create other splinters. Not just one.”
Claire asked, “How do we find them?”
Oliver said, “We look for the smallest changes that are most useful.”
Claire nodded. “Keys. Tokens. Scopes. Automation roles.”
Oliver said, “And we look for the places where people stopped looking.”
Claire felt the jar of stones in her mind. “Normal,” she said.
Oliver said, “Yes.”
***
That evening, an event occurred that proved the attack was not over.
At 6:27 p.m., Julian posted in the bridge, “Support received a message from Oliver instructing us to temporarily re-enable the assistant for partner tickets only. It sounds urgent. Is it real?”
Claire’s blood went cold.
Oliver replied instantly.
“That message is not from me. Do not comply.”
Julian responded with a single word.
“Understood.”
Connor posted next.
“Partner is escalating hard,” Connor wrote. “They are claiming we violated contract by disabling the assistant support integration.”
Rita replied, “Contracts do not override identity controls.”
Marianne wrote, “We need to handle the partner relationship carefully.”
Claire read Marianne’s line and felt the earlier forged sentence echo again.
Restore normal.
Handle carefully.
The attacker was trying to use executive anxiety as leverage.
Oliver wrote, “We are being actively impersonated. This is now an internal trust incident. All instructions to change systems must be verified through a voice call using a known number. No exceptions.”
Rita added, “And if the instruction involves access or automation, it requires my verbal confirmation.”
Claire saw the building’s ritual forming in real time. That was good.
But it also meant something terrible.
The attacker had enough visibility to know whom to impersonate and when.
They were watching the incident response and steering around it.
This was not opportunistic.
This was intimate.
Claire wrote in her notebook: attacker has eyes.
How?
***
Edward’s list of affected systems arrived at 9:02 p.m.
“Systems consuming the compromised rules repository,” Edward wrote. “Support assistant. Partner portal classifier. Internal template generator. And a small service called Gatekeeper that evaluates whether a request qualifies for time-bounded access.”
Gatekeeper.
Claire’s heart rate increased.
Gatekeeper was the system that decided which requests could be approved quickly.
It was supposed to be conservative. It was supposed to say no most of the time. It was supposed to be a wall that kept humans from approving dangerous things under pressure.
If Gatekeeper’s rules were compromised, then the attacker did not need to persuade a person as much. They only needed to persuade the system.
Claire wrote, “We need to audit Gatekeeper outputs.”
Edward replied, “Already pulling logs.”
Oliver wrote, “Disable automated approvals from Gatekeeper until audited.”
Rita replied, “Done. Manual only.”
Marianne wrote, “Manual only will create backlog.”
Rita replied, “Yes.”
Claire’s phone buzzed with a private message from Aisha.
“I think they aimed at Gatekeeper,” Aisha wrote. “If they can change rules, they can change what qualifies as safe.”
Claire wrote back, “That is the throat.”
Aisha replied, “Yes.”
Claire’s hands tightened on her phone.
The throat of privileged access was not only a human. It was a system. A system that the company had built to reduce friction. A system that had been fed the company’s desire for speed.
A system now potentially whispering yes when it should scream no.
***
At 10:18 p.m., Edward posted a screenshot.
Gatekeeper had approved an access request that should never have been approved.
The request was framed as verification.
It used a phrase similar to elevated visibility, but not identical. It used a new euphemism: “diagnostic window.”
The request had been approved automatically by Gatekeeper.
And it had been approved at 6:26 p.m., one minute before Julian posted about the fake Oliver message.
Claire felt a cold certainty.
The attacker had attempted to create a new splinter. They had used Gatekeeper to approve a dangerous window. Then they had impersonated Oliver to push Support to use it.
It was choreography.
Oliver wrote, “Who requested it?”
Edward replied, “Service account. It looks like it came from the partner portal integration.”
Connor wrote, “They have a portal integration that can request internal diagnostics.”
Claire wrote, “That integration should be isolated.”
Edward replied, “It is now.”
Rita wrote, “Any approvals granted by Gatekeeper in the last thirty days are under review. Revoke all time-bounded access grants until verified.”
Julian wrote, “That will lock out some legitimate cases.”
Rita replied, “Yes.”
Claire wrote, “Better to inconvenience the legitimate than empower the illegitimate.”
Julian replied, “I agree.”
That sentence warmed Claire more than it should have. It meant Julian was with her now, not only tired, but aligned.
Oliver wrote, “We need to find how the partner portal can trigger Gatekeeper.”
Connor wrote, “I can map the integration. I have contacts.”
Oliver replied, “Do it. But do not trust what they tell you. Trust what the logs say.”
Connor wrote, “Understood.”
***
The hunt moved from the build to the portal.
Connor arranged a call with the partner’s technical liaison for the next morning.
He asked Claire to join. He asked Oliver to listen.
Claire agreed, but she told Connor something before the call.
“We assume they are hostile,” she said.
Connor’s face tightened. “If I treat them as hostile, they will know.”
Claire replied, “They already know. They are inside our walls.”
Connor swallowed. “Okay.”
That night, Claire slept for two hours on the couch in the conference room. She woke with her heart racing. Not from a nightmare, but from a message. Her phone lit up with a notification from Marianne.
“Need to discuss partner strategy,” Marianne wrote. “Call me.”
Claire stared at the text.
It could be real. It could be forged.
That was the point. The attacker had turned communication into an uncertain terrain.
Claire did not reply.
She walked to Rita’s office instead, knocked once, and waited.
Rita opened the door, looked at Claire’s face, and did not ask for explanation.
Claire showed her the message.
Rita said, “We do not respond to text. We verify.”
Claire nodded. “I know.”
Rita picked up a landline and dialed a number from memory, not from her phone’s contact list. Her memory was a control.
Marianne answered.
“Yes,” Marianne said, irritated. “I messaged Claire.”
Rita said, “Claire is with me. We are verifying the request.”
There was a pause on the line. Then Marianne’s voice shifted, becoming more contained.
“I need to speak with Claire,” Marianne said. “In person. Now.”
Rita looked at Claire, then said into the phone, “Come to governance.”
Marianne said, “Fine.”
Rita hung up and looked at Claire.
“Do not go alone,” Rita said.
Claire nodded.
In a building where words could be forged, physical presence was beginning to matter again.
***
Marianne arrived ten minutes later, wearing a coat too thin for the cold. Her hair was pulled back, her face pale with fatigue.
She sat in the chair opposite Claire and Rita, and she did not waste time.
“This partner cannot be lost,” Marianne said. “They will go to press. They will say we are unstable.”
Claire said, “We are unstable because we were compromised.”
Marianne’s eyes flashed. “Words. We need to manage perception.”
Rita replied, “Perception is not a control.”
Marianne turned to Rita. “Governance is not revenue.”
Rita’s expression stayed flat. “Governance is survival.”
Marianne looked back to Claire. “We have to restore some normal function. We have to re-enable something.”
Claire heard the word normal again and felt a cold line form behind her eyes.
“We are being actively impersonated,” Claire said. “The attacker tried to use Gatekeeper to open a diagnostic window. They impersonated Oliver to push Support. Your partner is the pressure surface. We are not restoring normal. We are restoring safety.”
Marianne leaned forward. “If we do not provide the partner with a path, they will leave. That is real damage.”
Claire replied, “If we provide them with a path that can be exploited, that is existential damage.”
Marianne’s voice tightened. “You always assume worst case.”
Claire said, “No. I assume exploited case. We have evidence.”
Marianne pressed her lips together. “So what is your plan? You cannot just say no.”
Claire pulled up Oliver’s message in her mind. Make it boring. Make it procedural.
“My plan is to create a verified partner support path that does not require any access escalation,” Claire said. “No diagnostic windows. No time-bounded privileges through partner portal. No Gatekeeper triggers. Human handled, identity verified through secure channels, and every request logged.”
Marianne frowned. “That will be slow.”
Claire replied, “Yes.”
Marianne said, “They will hate it.”
Claire replied, “Then they can prove they are a partner by tolerating safety.”
Marianne stared at her.
Rita added, “And if they refuse, we learn from that too.”
Marianne exhaled sharply. “Fine. But we need a message. We need to tell them something.”
Claire said, “We tell them we are performing security hardening after detecting manipulation attempts. We offer a safe support channel. We refuse unsafe automation. We do not apologize for boundaries.”
Marianne’s mouth tightened. “We cannot accuse them.”
Claire said, “We do not accuse. We state facts. We state that we detected manipulation attempts. We do not name them. We set terms.”
Marianne’s eyes flickered. “You want to set terms to a partner?”
Claire replied, “Yes. Partnerships are not ownerships.”
Marianne stood, a burst of motion that looked like anger trying to become authority. Then she stopped, as if she remembered she was in Rita’s office and Rita’s office was the kind of room where authority had to be verified too.
Marianne said, “Draft it.”
Claire nodded. “We will.”
As Marianne left, Rita looked at Claire.
“She uses normal like a drug,” Rita said.
Claire replied, “And the attacker knows it.”
***
The next morning, the partner call began politely.
Connor joined from a conference room. Claire sat beside him. Oliver listened silently,
off camera, his mic muted. Rita joined as an observer. Marianne did not join. Claire had insisted. Too many voices made the partner feel powerful.
The partner liaison introduced himself as Caleb. His voice was smooth. His tone was sympathetic.
“We understand you have had some internal disruptions,” Caleb said. “We appreciate how hard your team is working.”
Connor nodded. “We are doing our best.”
Caleb said, “Our concern is service continuity. We rely on your platform.”
Claire listened, not to the words, but to the structure.
Praise.
Concern.
Reliance.
Then, inevitably, the lever.
Caleb continued, “We also rely on the diagnostic tools that have helped us in the past. We would like those restored.”
Connor began to speak, but Claire raised a hand slightly, a silent signal.
Connor stopped.
Claire said, “Which diagnostic tools?”
Caleb replied smoothly, “The ability to open a verification window when needed.
It speeds resolution.”
Verification window.
New euphemism.
Claire said, “We have disabled automated diagnostic windows because we detected manipulation attempts. We can support you through a verified path that does not require privileged access.”
Caleb’s tone stayed pleasant. “We have never manipulated anything. We only want support.”
Claire said, “We are not accusing. We are stating that we detected attempts to steer internal systems. We are hardening. Our path is the safe path.”
Caleb paused, and in that pause Claire heard something she had not heard before.
A calculation.
Caleb said, “We have contractual obligations. You are obligated to provide reasonable support mechanisms.”
Claire replied, “We are providing support. We are not providing unsafe automation.”
Caleb’s voice cooled slightly. “If you cannot provide the diagnostics, we will have to escalate.”
Connor swallowed. “We can work with you.”
Claire watched Connor’s impulse to soothe. It was a habit. A habit the attacker could exploit.
Claire said, “Escalation is your choice.”
Caleb laughed softly, as if Claire had said something naive. “We are partners. We want collaboration.”
Claire replied, “Then collaborate with boundaries.”
Caleb’s voice tightened. “This is not acceptable.”
Claire said, “Then you have intelligence to deliver to your leadership. We are safe.
We will remain safe.”
Caleb paused again. Then he said, “We will follow up in writing.”
Connor said, “Okay.”
The call ended.
Connor looked at Claire as if he had just watched her break a vase he had been trying to keep intact.
“You were harsh,” Connor said.
Claire replied, “I was clear.”
Connor said, “They will escalate.”
Claire said, “Good.”
Connor stared. “How is that good?”
Claire replied, “Because now we can see their intent.”
Oliver unmuted his mic. “And because their escalation will create pressure that reveals their influence paths.”
Connor looked toward the speaker. “You were listening.”
Oliver said, “Yes.”
Connor exhaled slowly. “Okay.”
Claire looked at Connor. “You did well. You did not offer a door.”
Connor’s mouth twitched with something like pride and fear mixed together.
“I wanted to.”
Claire nodded. “I know.”
***
Within two hours, the partner escalated.
The escalation did not come as a threat. It came as a complaint written in careful legal language. It came with a list of “service expectations.” It came with a demand for restoration of “diagnostic capabilities.”
And it came, crucially, with a portal link.
The portal link pointed to a support integration endpoint that Connor had not mentioned. It was a hidden piece of the integration, a backchannel used for “premium diagnostics.” It was allowlisted in the company’s internal network.
Allowlisted.
Claire hated the word. It sounded like purity. It was really just a door with a polite sign.
Edward traced the endpoint.
“It can submit requests to Gatekeeper,” Edward wrote. “That is how the diagnostic window got approved.”
Oliver wrote, “So the partner has a direct lever into our access evaluation system.”
Rita wrote, “That lever has to be removed.”
Edward replied, “We can remove it, but it will break the partner integration.”
Marianne wrote, “We cannot break the partner integration.”
Claire read the message and immediately felt suspicious.
Was Marianne sending it? Or the attacker using Marianne’s voice?
Claire did not respond in the channel.
She walked to governance, where Rita was, and asked Rita to verify Marianne’s statement through a call.
Rita called Marianne. Marianne confirmed she had written it.
Claire felt anger rise, sharp and clean.
She returned to the bridge and typed carefully.
“We will have to break the integration,” Claire wrote. “We must break the lever, there is no choice.”
Marianne replied, “That will create revenue risk. There is always a choice”
Claire fired back, “Not this time.”
Rita replied, “Revenue risk is acceptable. Compromise risk is not.”
Oliver wrote, “Disable the endpoint. Now.”
Edward replied, “Executing.”
At 2:17 p.m., Edward posted, “Endpoint disabled. Partner premium diagnostics broken.”
Connor wrote, “They will lose their mind.”
Claire wrote, “Let them.”
The hunt had become overt. The attacker would notice their lever was gone.
That meant the attacker would move.
Claire wrote in her notebook: once lever removed, expect retaliation.
