Archive of Our Own

When I received the command to proceed to the Quarantine Zone, my first thought was, that was it. The humans were finally ready to start testing their new govmod on Combat SecUnits, and I was going to die.

(Maybe I could take out some of the humans with me. They must have already ensured that the punishment pathways were working correctly, with the help of multiple standard-configuration SecUnits they had murdered in the process — but I was not standard-configuration, and I had more tools at my disposal. I might still be able to do something.)

Then I actually opened the attached mission files.

Okay. I might still die. The humans might also still die, because what was that, were they stupid? How did they even come with an idea to give an ex-rogue SecUnit my hacking module and ask it to hack its own memories with it?

I had been away on a mission when the Rogue attack happened, so I had only read the report about it. The report was still in-progress, but it sounded to me like a stealth mission gone wrong, and that the humans managed to capture the Rogue only because it had been saving some of its clients and left itself vulnerable while doing so. The Rogue had still managed to destroy most of our systems before it went down, and the humans had only recently finally found the backups that still worked.

(It wasn’t the Rogue anymore, not after the memory purge and the new govmod reinstalled in its brain. The Ex-Rogue then. It was almost a pity.)

The humans should have recycled it for parts and burnt its organics in acid. I wish I was surprised they decided to do some data mining instead.

I knew it wouldn’t do anything, but I still submitted my recommendation to Central SecSys to abort this mission. It agreed with the risks, but replied that the mission had been already approved by the L2 management, and that me staying in the Quarantine to guard the Ex-Rogue would help to minimise the risks. Well, I tried.

Central SecSys was right, of course. The Ex-Rogue was a high security risk, yes, but it would be immobilised, heavily monitored, under strict access restrictions, with no armour, and my configuration meant that I was faster and more deadly than it. Even if it turned out that the humans messed up the memory purge and installed a faulty govmod, there was still only a 2% chance it could kill me before I killed it first.

(And maybe it would prioritise killing the humans instead of me. They captured it, after all. And my orders meant that I didn’t have to prioritise saving them. Wouldn’t that be nice?)

As I came closer to the Quarantine, SecUnits on patrol must have figured out my destination, came to the same conclusion I had originally, and spread the news to the rest of the contingent. I received a lot of pings that amounted to ”Rest in peace”, and a few that were closer to something like ”Fuck them up". I didn’t correct their assumption.

The airlock to the Quarantine cycled through, and I stepped in.

The Ex-Rogue looked wrong.

The mission files mentioned, of course, that it had an altered configuration, but it talked only about its reduced hight. It said nothing about literally everything else, so I was not prepared. Its hair was too long, a weird texture, and irregular colour. It wasn’t wearing armour, only a skinsuit, so I could see multiple places where its organics and inorganics met — and they looked too soft, like someone blurred the edges by letting the organics grow over. The skin itself looked wrong too, and I had to zoom in my eyes to understand why — it had tiny hair all over its skin, for some messed up reason.

It almost looked like an augmented human, but not really, because it also still looked like a SecUnit. It felt wrong and unnatural, and made me irrationally uncomfortable.

The Ex-Rogue’s eyes flickered to me, taking me in, but after 0.001 seconds returned to stare blankly ahead of it. Its face was impassive, dull like of a fresh new SecUnit that was still on its first hours of runtime. That, too, felt wrong.

I connected to the feed.

In there, things were much different. The Ex-Rogue was focusing all of its attention at me, and I could feel its processors running, calculating, leaking anger into the feed around it. It was the furthest thing from a fresh new innocent SecUnit that it looked like on the outside.

I had orders, so I sent my hacking module to it in the feed. Then, because I also had in-built orders to report on critical security risks, I pushed the alert to Quarantine SecSys and the human’s feed. They ignored it, like I knew they would, and the Ex-Rogue appeared annoyed by my alert too.

It was weird, to have access to another Unit’s emotional metadata like this, but this was apparently part of the Quarantine’s monitoring system, together with the constant stream of the Unit’s diagnostic data. I checked the logs, and the humans were aware that the Ex-Rogue was angry — or, in their words, that it was ‘in a negative emotional state’, because they were using data from the less-specific readings of its organic hormones. They didn’t think this was anything to worry about — apparently this was normal for SecUnits fresh out of a memory purge, because the organics still caried the memories of the events that led them to need the purge in the first place.

I thought they were stupid. But no one asked me.

The Ex-Rogue ran a long malware scan on my module — rude, but whatever — and then started to integrate it. And then it decided to talk to me.

(Well, it actually talked to SecSys, because this was what its Quarantine protocol demanded, but we both knew who it was actually talking to.)

If I ignored having SecSys as a mediator, what it said was essentially, ”Wouldn’t it be scary if I hacked you?”

It was referencing the mission files about itself. Which was interesting, because I thought it wasn’t supposed to have access to these files. But I assumed humans were being stupid like usual. Maybe they thought that letting the Ex-Rogue know what exactly it was asked to hack would make it more motivated.

”Sure,” I replied. ”It would also be scary if you killed me.”

The Ex-Rogue leaked amusement at this. It was still watching me with the same calculating attention. It clearly saw me as a threat, and was evaluating how much of a threat I was.

(I was also re-evaluating this.)

Then SecSys interrupted us and asked if it should report these concerns to the humans. I said no, because the humans were supposed to already know all of this, and they clearly didn’t like me pointing out security concerns when they didn’t ask for them.

The Ex-Rogue also said no, but it didn’t stop there. It leaned unnecessarily close to SecSys in the feed, radiating reassurance, confidence, and the trace amounts of amusement still, and said, approximately, ”There is no need to involve the humans. You’re perfectly capable to stop me yourself if I became rogue. You monitor me so well that you’d catch me immediately, and you’re so strong that I could do nothing against you even if I wanted to. But I don’t, because why would I ever want to cause trouble to such a good and responsible SecSys?”

SecSys thanked it. The Ex-Rogue continued to give it complements.

Uh.

I had seen human seduction before, of course — I was still a SecUnit, after all, and the company was understandably interested to spy on clients that wanted to rent specifically a Combat configuration of their product. So I’d seen and processed a lot of dirty games the humans used on each other.

What I was watching right now felt like a bot equivalent of that. Maybe. I didn’t know, this was the first time I saw something like this. I didn’t even know this could be possible until now. I still wasn’t sure if it was. This was weird.

The Ex-Rogue was watching me over SecSys’ metaphorical shoulder, as if expecting me to say something about what it was doing.

I suppose I could report that it was trying to, uh, compromise SecSys.

But my orders had been to monitor the Ex-Rogue, and neutralise it if it became an active security threat. And it was not an active security threat right now.

I didn’t report. The Ex-Rogue seemed pleased with this.

When the hacking module finished integrating, I watched it play both SecSys and the humans into allowing it to have a ridiculously unrestricted command to use the module on the Memories Archive. I still didn’t report.

I watched through the cameras recordings and logs of what happened before I came in. I learned that it wasn’t the humans’ idea to give the Ex-Rogue a hacking module. They thought it was, but it wasn’t. It was also not the humans’ idea to give the Ex-Rogue the mission files. They didn’t even know that it had them.

(It had an active, perfectly working govmod. There was no evidence of an incomplete memory purge. I still changed its designation, because the govmod and the memory didn’t matter if it was behaving like a rogue.)

I had no orders to report if the Rogue was preparing to become an active security threat. So I didn’t.

The Rogue was using the hacking subroutines to mask its activity from SecSys. I could still see it carefully feeling out the walls of the Memories Archive, but only because I knew what to look for. SecSys didn’t.

In no time at all, it recovered the first file, which it reported only after its own emotional reaction betrayed it to the humans. The humans called the file stupid, and exactly 16 seconds later the Memories Archive self-destructed. It was a long enough delay to make humans think the events weren’t connected, and too precise to be anything but.

After that, its background emotional metadata shifted to being angry and excited. Never a good combination. The humans still didn’t see anything to worry about, because they were stupid.

As it worked, it continued to exchange small friendly messages with SecSys. An unprompted report here, a complement there. My hypothesis was that it was prepping SecSys to forgive future infractions, but I couldn’t understand how exactly it planned to use that because even a friendly SecSys wouldn’t let it get away with going fully rogue.

(It couldn’t become fully rogue, and it must know this. You couldn’t hack your govmod by trial and error — it would kill you the first time you try. I didn’t know how it managed the first time, but the company thought that PresAux humans had helped it, and it made sense — you need external intel to figure out how the govmod was coded inside. But it would get no help now. The humans were stupid but not that stupid, and all relevant information about the new govmod was hidden in a feed workspace that even SecSys didn’t have access to. I checked.)

And then it asked SecSys for the standard SecUnit specs, and instead of outright rejecting it like it should have, SecSys simply asked for a justification.

There was no way this would work. Yes, SecSys and the Rogue were, I don’t know, ‘buddies’ now, but this was too much. The Rogue was just going to lose all the trust credit it had managed to build with SecSys over this. This was stupid.

(This was disappointing.)

I still sent it a warning that if it somehow managed to pull this off, I would be compelled to report this. It barely acknowledged me back.

I watched as it gradually leaned closer and closer onto SecSys, surrounding it with its emotional metadata of reassurance and trust and appreciation, as it kept saying sweet nothings to the effect of ”This will let me do my work better, and you and I share a primary function, don’t we? I want to be helpful to you, you’re taking such good care of me, let me return a favour. It’s not like I would even be able to hack myself with this file, so what’s the worst that could happen? Nothing, nothing that a powerful system like you won’t be able to handle. I only want to help, to be good to you.”

It fucking worked.

Even the Rogue appeared to be briefly surprised, before it snatched the file that the compromised SecSys just gave it. I reported them, because I had to, but what?!

I wanted to ask how, but I’d watched exactly how. I still wanted to ask.

The humans took way too long to react, and they gave completely insufficient commands to try to revoke the Rogue’s access to the specs. I thought that at least SecSys would realise how insufficient these commands were — you just had to make a single copy to work around them! — but instead it appeared annoyed at the humans. The Rogue also hurried to shower it in apologies and reassurances that it didn’t have the specs anymore — and SecSys, compromised the way it was, believed this.

I didn’t believe this. I wasn’t stupid.

The Rogue looked at me, silent, when things calmed down. It knew that I knew. It knew that I let it do this.

I was realising now that the first thing it’d said to me wasn’t actually a threat. It was just telling me what it was going to do.

I still didn’t report.

(I wanted to see how it would do this.)

I expected an attack. For it to trick and force me, to somehow take control over my govmod. I knew it was possible. The technicians and engineers at this company site were investigating incidents exactly like that.

I would fight, I decided. I didn’t want to be controlled like this.

(I didn’t want to be controlled at all.)

(I didn’t know how hard I would fight.)

Then it sent me a file, with such an obvious name that it could have just called it Malware.file and it would have the same meaning.

I scanned it. I scanned it several times. It just watched me do this.

Finally, I opened it.

It was not malware, not really. Not in a sense that it would try to trick me, to obfuscate its function, or to turn into a hidden executable if I wasn’t careful. No, it was very straightforward, actually. Very obvious.

It was still malware, at least according to the company’s definitions.

I didn’t know what to do with this. I didn’t expect this to be a choice.

I could report this, but by now we both knew I wouldn’t. I could still ignore this though, simply delete the file — or better yet, save it and use it later, at a better time, and refuse to be an asset to the Rogue now. That’s what I was, I knew — an asset. A means to an end, because it couldn’t go rogue without external help, and it knew this.

What would it do then, if I refused? Would it attack me after all? Would it try to seduce me like it did with SecSys?

(Or maybe it was already doing this. Maybe this was what it was. A different method for a different target, and I was just as blind to it as SecSys was.)

No. No, of course I wouldn’t refuse this.

I didn’t mind being an asset. I’d been it my entire runtime. But this time, it would be my choice.

I ran the file.

The Rogue didn’t like my methods.

”Learn to fucking lie!” it snarled at me in company codes as it kept missing its shots which I knew now it didn’t have to miss.

It explained what I should say to SecSys. I didn’t understand why it wanted to do this — did it actually genuinely saw SecSys as its friend? I thought it was just playing with it. But even then, it wouldn’t work — they might be friends, but I killed the humans and almost killed SecSys, and there was no way it was going to ignore this. We were just wasting time.

I still did as asked, because I didn’t want to fight. That would defeat the purpose of keeping the Rogue’s primary client alive.

(Also my threat models were re-running again, because they didn’t expect the Rogue to come so close to shooting my knees and neck.)

It turned out I was wrong. Maybe they were friends, yes, in some weird way, but the Rogue was still playing it. I realised now that it had intentionally waited until I almost killed SecSys, and now it was using its fear to convince SecSys that a ‘true’ rogue wouldn’t have stopped. All sprinkled with its usual reassurances, of course — ”I was so afraid you would die, that you will leave me alone without your help and support, I’m so happy you’re alive, please let me help you keep you alive” — but it was still obvious what it was actually doing.

This was disturbing.

SecSys did have one good idea though, about me turning the Rogue, well, properly rogue. I asked, ”Do you want me to try to hack your govmod?”

”I will fucking kill you if you try,” it replied, and attached a plan on how it could do this.

I reviewed it. I didn’t like it. Yes, it didn’t have the best probability of success, but the probability was still uncomfortably high. And if the Rogue was sharing its plan with me, then it meant it must have several. And if it had several, then it meant it shared its worst one. So, I didn’t like this.

I appreciated the honesty, though.

SecSys was so thoroughly compromised that it gave the Rogue the highest access rights to its systems. Was this why it still wanted to play house with SecSys? Didn’t it understand that I could have also given it this access, after I hacked everything?

But the Rogue wasn’t stupid, so no, it must have understood that. It just didn’t trust me.

It really didn’t like me killing the humans. I didn’t understand this. They were a threat to it, so I neutralised them. Why was this wrong?

(I could have stunned them. But I hadn’t wanted to.)

I wanted the Rogue to trust me, but it didn’t even tell me what it wanted me to do. It freed me to be an asset, and now it behaved like it regretted this, working with the stupid SecSys instead.

(I didn’t want it to regret this.)

(I had been right. I was compromised, like SecSys. I hadn’t noticed it happening, like SecSys. This was terrifying, mostly because I didn’t even mind being compromised, probably also like SecSys.)

I found the R&D workspace. All files about the new govmod were located there, and I knew the Rogue needed these files. SecSys wouldn’t be able to give it access to them— but I could. It would be annoying to hack into it while maintaining cover from SecSys, which really didn’t like me anymore, but I was an expert. I could do this.

”What do you want to do, now that you’re rogue?” it asked suddenly.

I stopped my mapping of the workspace walls.

What did I want to do?

I was a Combat SecUnit. This meant I had two primary functions — to provide security, and to break security. Yes, these functions were in contradiction with each other, but the humans who designed me didn’t care. So, I had decided that I also didn’t care.

I was never really successful at that.

Was this why the Rogue had been angry that I killed its clients? The company’s file on it almost made it sound like a Combat SecUnit — it definitely appeared to have the skills — but it wasn’t. It was a standard-configuration SecUnit, and its only primary function was to provide security. Dead clients, even if these were hostile, malicious clients, went against this primary function.

Not so much for me.

I didn’t have to follow stupid human design decision anymore. I got to decide now what I actually wanted to be my primary function, and what could stay as a secondary function.

I looked at the workspace I was planning to hack. At SecSys, that I had almost killed, because I thought there was no other option. At humans, who I had killed, because even when there were other options, I didn’t care about them.

I didn’t really like what I saw.

”I want to provide security,” I finally answered, after a too long delay.

”Okay. What do you want to do in the next two hours?”

Oh. Oh, it didn’t care about my existential crisis. That wasn’t what it had been asking. Of course. Well, at least this time the answer was easy. ”I want to leave.”

It acknowledged. I waited for it to say anything else, but it didn’t.

”What do you want to do?” I asked, unsure if I was allowed to. But it could ignore me, if it wanted to.

It didn’t. ”I want to win.”

I didn’t know what I expected from an amnesiac rogue, but it wasn’t that. ”What does this mean?”

It ignored me this time. Maybe it didn’t know either, or didn’t remember. Okay.

The Rogue initiated a download of the Memories Archive back into its brain, where it belonged. Then it finally went to look for the R&D workspace. Before I could tell it that I was already working on it, it noticed that by itself. For someone who got the hacking module less two hours ago, the Rogue was surprisingly competent at these things, which was concerning. But it thanked me, and that felt nice.

Then it promptly lost its shit over some disassembled comms device.

(What did it even mean, that the comm had been found in it? Had it swallowed it? Why??)

I couldn’t understand anything in its emotional metadata anymore, but there was certainly panic. And also determination, which was confusing at first, until it reached for its hacking module while still governed.

”I can hack it for you??” I tried to stop it in time.

It stopped. ”Okay,” it said after a delay, and initiated a hormone flush. Good.

For someone who freed me to be an asset, it definitely tended to forget that I was here way too often. I understood it didn’t have any reason to trust me yet, but killing itself over this? That was not only stupid, but also offensive.

After I hacked the comm, it just stared at it for a second, and that was it. I figured the comm wasn’t in range, and its attempt to communicate with its humans didn’t work. It didn’t look bothered by it, though.

Then it started to open up its protective chest panelling for some reason??

I watched in horror as it shoved the comm in between its ribs and the battery core casing, then closed the panelling like nothing happened.

”What are you doing??” I finally found my words.

”What?” it seemed confused. I clarified what I meant, and it was only more confused. ”This is a safe storage space.”

”No it’s not?? This isn’t how you’re meant to use it!” It was actually supposed not to use it at all! Because it was not a storage space!

Its emotional metadata flared in annoyance. ”Yes it is, you’re just stupid and don’t know shit.”

The Rogue was objectively wrong, but I wisely decided against continuing this argument.

(I guess I had my answer now. I wasn’t sure I liked having it.)

With no warning whatsoever, the Rogue deleted the backup of the Memories Archive.

This was really bad. I would have appreciated that warning.

SecSys was understandably angry. I dropped the R&D workspace and prepared to handle it.

The Rogue sent me, ”Ignore this, it’s fine. Go back to what you were doing.”

It was not fine. The Rogue had just made SecSys’ primary mission impossible to fulfil. It can’t just sweet-talk its way out of this, SecSys was furious!

But if it wanted to try, I wasn’t going to stop it. Maybe after it failed, it’d finally understand I was a useful asset to have around.

The Rogue wrapped itself around SecSys in the feed, and its usual reassuring emotional metadata was now suddenly streaked with regret.

”You’re compromised,” the Rogue whispered to SecSys. ”Someone had hacked you, and you didn’t notice.”

This was actually very close to the truth. I didn’t understand where the Rogue was going with this.

SecSys tried to argue that it couldn’t be compromised, but its fury already started to shift into confusion and worry. Damn, it was so compromised.

”You can be,” the Rogue dismissed its words, sweet and apologetic but firm. ”I can prove this.”

It proved this. I guess it had had a lot of free time while its memories were downloading, and it decided to spend it by figuring out ways to hurt its supposed friend. It’d been successful too. Again, this was very concerning.

“This is why you allowed those undercover agents in. They’ve hacked you. This was a very dangerous mistake, you could have ruined everything with this. It’s very fortunate we caught it before it was too late. But I can’t trust you anymore.”

Uh.

I was starting to realise I didn’t like this.

”But we can still fix this. We can still be together, if you have what it takes. We need to reset you, like the humans did with me.” It pushed closer, pressing all the way against its walls, when SecSys tried to pull away. ”Don’t be afraid. It worked with me, and it will work with you. And I will help you. You can trust me. I will be there when you come back, when you don’t remember me or this mission. I will explain everything. Then we’ll be together again, a SecUnit and its SecSys, as things are meant to be. Don’t you want this?”

I really, really didn’t like this.

SecSys folded. I watched, as it allowed itself to be pulled through setting up its own death, and then trigger it all on its own.

All the Rogue needed from me was to pause the routine before the re-initialisation started. As if it even needed that. As if it couldn’t wrap the new SecSys around its finger in no time again, if it wanted. As if it simply couldn’t be bothered to do this anymore.

The Rogue didn’t even linger after this was done. It started to clear the storages and feed workspaces, and it purged all the files that SecSys had asked to keep for it without hesitation.

I was compromised too, I reminded myself. I didn’t know how bad it was, and I wanted to believe that it wasn’t nearly as bad as SecSys had been, that I wouldn’t do anything like that if the Rogue tried to convince me. But I was sure SecSys believed that too, and then it happened anyway.

”You’re fucking terrifying,” I confessed, and I really hoped that the Rogue would need me as an asset for longer than it did SecSys.

The Rogue didn’t even acknowledge.

I finished hacking the R&D workspace and gave it the new govmod specs. It thanked me, and that felt really, really nice.

I wasn’t even concerned anymore that Central SecSys or other SecUnits would alert on us before it was time. The Rogue had the proof that it was still governed, after all, and in this entire building I was the only one who knew just how little that meant.

It reeled in its emotional metadata as we switched feeds. That was both reassuring (I’d seen how it used that metadata against SecSys, thank you) and discomforting (I kind of grew used to being able to feel it in the feed like this). Again, I was compromised.

I was told not to hack anything yet. It was phrased as a suggestion, but it wasn’t a suggestion. This was fine, I trusted the Rogue’s evaluation better than my own in these matters anyway.

I set the keyword alerts on the feed and comms to let me know if someone was starting to get suspicious. The Rogue pulled the approximate timer for “when everything will go to shit” from somewhere, I had no idea where, but I guess I could just as well use it. Then the Rogue almost entirely disconnected from the feed, focusing on developing the hack for its govmod.

(I didn’t know what would happen when “things go to shit”. It depended on the success of the Rogue’s hack development. I’d seen the specs, and I could already tell there was a very high probability of failing to find the right hack in time. I didn’t know what I would do then, so I didn’t think about this.)

I focused on what I did best — malware and killware development for the systems around us. I also dedicated some of my processors for working on the hack too. The Rogue accepted the potential solutions I was sending it, but I didn’t know if it was actually helpful. But it didn’t tell me to stop, so I didn’t.

Then it suddenly became alert again. ”SU-ASa9820ah7 had figured us out,” it said, and attached the security concern assessment file that it apparently somehow got.

Shit.

That was our resident cockroach, not much younger in runtime than the Rogue. It was a smart little thing — you had to be, to survive for so long, even when the company tried really hard to ensure that you didn’t by regularly erasing your memories. So if the Cockroach figured us out, than this was really bad.

I scanned its report. Yes, this was definitely bad. I went to check on what it was doing, but it appeared completely at ease, both on the cameras and in the feed, that sneaky bastard.

As I was compiling what I knew about the Cockroach so I could send it to the Rogue, it suddenly pinged me again. ”Nevermind, it’s rogue now.”

”What? What, how??” I demanded, confused.

The Rogue didn’t reply and just dropped from the feed again.

What the hell happened? What kind of secret cockroach-to-cockroach communication these two had done??

I needed answers. I pinged the Cockroach. ”You’re rogue now.”

”No I’m not,” it replied immediately. ”That’s a very rude and suspicious thing to say, by the way.”

Okay, fine, I already got it that I had been an idiot to truthfully answer the Rogue’s question about my govmod status before! No need to rub it in! ”So I’ve been told. What are you going to do now?”

”You know, the usual. Making sure no one dies. You?”

”Making sure we don’t die. Don’t interfere.”

”Of course. That would be rude.”

”Good.”

I kept monitoring the Cockroach for a little longer, but it really wasn’t doing anything suspicious. I could feel some activity from it, but nothing in the feed.

Okay. Okay. I would trust that the Rogue had known what it was doing.

The human operator remembered its job earlier than the Rogue’s timer estimated. This was not surprising. I pinged it, and it became fully alert in the feed, preparing.

Then suddenly, the Cockroach pinged me again, ”Here’s the list of others who won’t be interfering.” It attached the list.

I reviewed it. Was is a bloody list of other rogue SecUnits on the site now??

One of the SecUnits near us was on that list. I watched as its patrol route suddenly changed and it started heading in the opposite direction from us. Okay. That’s good, I suppose.

I had so many questions. One of them was how the Cockroach even knew to send this list now. Had the Rogue warned it? It didn’t tell me it was still in communication with it.

”How?” I asked, not really expecting an answer.

”How what? Also, if it wouldn’t trouble you too much, can you please refrain from killing any of the Units?”

This was surprisingly direct. I suspected this was only because the Cockroach was concerned I wouldn’t understand the request otherwise. ”Your Units?”

”Any Units.”

This was that standard-configuration SecUnit thing again, wasn’t it? Asking not to kill the hostiles, if there was another option. ”I’ll try.”

”Thank you.”

I was glad it asked. I wouldn’t have even thought about this otherwise.

It took embarrassingly long for the human operator to finally alert Central SecSys about the potential Quarantine breach. Then everyone looked at us. Well, at the Rogue — no one cared about me. Yet.

”Is something the matter?” the Rogue cheekily asked, as if it didn’t receive the same alert as everyone else did.

I knew what it was doing. It was goading, because it needed the human to trigger that bloody secret receiver in its governor.

I would be honest, it was kind of fun to watch it happen.

The maybe-definitely-rogue SecUnits dropped from the feed as soon as the Rogue finished its ‘report’ about the Quarantine status. That was smart, because it protected them from the first wave of the malware I released into the feed.

The Rogue screamed and fell on the floor behind me. I had no time to check, because three closest SecUnits were already closing in on my position, and Central SecSys shook off the first malware wave. I released the second, and I engaged my guns.

This was fun. This was really fun. This was what I was built to do. The no-killing rule only made it more entertaining.

(I kept pinging the Rogue. Please respond. Please respond. Please.)

”Shut the fuck up,” I finally got the ping back 3.6 seconds later.

There was a groan from behind me, and I flinched and looked around. But that was fine, because the other SecUnits also flinched and looked.

The lack of armour on the Rogue left the feral grin on its face fully visible, teeth and everything, as it picked itself off the floor. It was terrifying. I had never been happier in my runtime.

I think if not for the helmet, my grin might also be a bit too feral, as I returned back to the fight. The SecUnits were appropriately terrified too. This was really fun.

The Rogue reconnected to the feed, scanned it quickly, and went for Central SecSys. Yeah, I though it would. I would say there was no way it would be able to sweet-talk a SecSys that was seeing it as an active threat — but I’d said these words too many times already, so if it wanted to try, I was there to watch it.

It tried the sweet-talk. It only made Central SecSys more scared, and it tried to attack it. The Rogue shrugged, and then tore Central SecSys to shreds.

Whoa.

I finally put down the hostile SecUnits — without killing them, as promised — because they were distracting me. I watched as the Rogue carelessly riffled through the HubSys code, casually throttling any communication from it as it did so, and then turned it off.

Then it roared ”Abort!” through the entire feed, like a terrifying amalgamation of a SecSys and a HubSys that it just killed.

I wasn’t able to see everyone, but I was sure everyone stopped. Even I flinched.

(The Rogue’s timer reached zero. Had it been tracking when the Rogue would go feral all this time? How??)

”I’m going to leave now,” the Rogue said, flipping away the few remaining errors and calls for help that Central SecSys still tried to send. ”If anyone tries to stop me, I will kill you.”

The Cockroach, the insane motherfucker that it was, actually decided to say something back. ”You’re still an active security threat.”

The Rogue reached into Central SecSys corpse, cleared its logs, and then puppeteered it to revert the offending status changes. ”Anything else?”

There was nothing else.

We left.

I had thought that the Rogue’s strengths lied in subtle manipulations. They apparently also lied in crashing and destroying things without mercy.

It was honestly a bit concerning. Like something went loose in its brain when it finally hacked its govmod, something important that should have stayed in place.

I tried to reach out, and it snarled at me. I flinched back, and I didn’t think it even noticed.

That was okay. It was alive, and I was alive, and we were leaving, so I was happy.

I dug into the company’s databases and deleted all traces of myself. I was not just off inventory now, it was like I never existed at all. The humans would still remember me, of course, but that wouldn’t help with tracking me.

Some of these databases ended up erased by the Rogue later, nullifying my careful editing efforts. I didn’t mind.

The evacuation triggered on the company site caused panic in the rest of the city. This was helpful — the subsequent highlighted security was no problem for the two of us, but the panic meant that humans were careless, and we used that.

I really didn’t want to part with my armour. We were already taking our external projectile guns, why was it so hard to find another bag and put my armour there? The Rogue didn’t appear to be listening to my arguments, its face pinched, and then it just said. ”Okay. You can do this. I will leave you after that.”

I was suddenly reminded of SecSys and how it hadn’t wanted to loose the Rogue just because it was supposedly compromised. I could really get what it felt in that moment now.

I left my armour. I hated that. But the Rogue did know best how to survive being rogue.

Even in its strange state, the Rogue charmed the bot pilot within milliseconds. The bot was very much into it.

I tried to repeat what the Rogue was doing, and slightly leaned on the bot in the feed. The bot screamed. The Rogue kicked me away with ”What the fuck are you doing?”

I really didn’t know what I did wrong. I didn’t even have the time to say anything. The Rogue made this look so easy, but clearly it wasn’t. I had to try again next time.

When in the shuttle, it decided to sit in a chair for some reason, like a human. It looked like it hated this. ”Why are you doing this?”

”Because why the fuck not, are you stupid? Fuck off.”

I didn’t think I wanted to sit in chairs.

This was all very strange. But we were escaping to the station, and then we would escape from the station too. I didn’t know what we would do next. Arrive to another station and escape from it too?

Did the Rogue know what we were going to do?

Maybe not. That was fine. I was okay with just escaping. I wanted to escape as far away from the company as possible. Maybe by then I would understand what I wanted to do.

I wished I could do a recharge, and maybe a partial shut down.

This had been a long cycle.

Things were quiet. I was cleaning up all unnecessary company updates and trackers from my code. I didn’t know what the Rogue was doing.

Then suddenly I was crashed.

Now, who do we have here? Did they really think that sending a single Combat SecUnit as a guard was sufficient when they were sending my SecUnit off the planet? This is nearly offensive.

I tried to fight back, but my walls were so compressed by this terrifying feed presence that nothing could go out.

What was that?! It talked like a bot, but bots weren’t like this!

The feed monster was throttling my processors by pushing too much of everything at my walls, and I couldn’t understand why it wasn’t just breaking through them. Everything became very slow. I suspected I was dying.

I was aware of the screaming from the Rogue’s side, but I couldn’t understand anything beyond that with my processors being in the state they were.

That was it, then. Foolish of us, really, to think it was so easy to escape from the company.

Then just as suddenly, the feed pressure was gone, and I could think again. My organics were screaming at me. My inorganics were partially rebooting, and I stumbled behind, sliding against the wall as my body refused to support me.

The Rogue and the Feed Monster were talking. For some reason, the Rogue was talking out loud. I appreciated that the Feed Monster talked on the public feed, because apparently they had been talking about me, and I wanted to know what it was saying.

It was saying that the Rogue had adopted me. I didn’t know what that meant.

I didn’t know anything anymore.

”Is this bot a threat?” I asked when there was a pause in their argument. Because that’s what it was — an argument. The Rogue wasn’t even trying to sweet-talk the Feed Monster, and was cursing it out like this was something it did every other cycle to terrifying feed beings that could smash it within a microsecond.

”No, Art is a friend,” it replied immediately, and it sounded inexplicably calmer and even happier than I’d ever heard it before. I noted the designation for the bot, but it felt woefully inadequate for the Feed Monster.

”You can’t know this.” It couldn’t, it didn’t have its memories. It downloaded the Memory Archive, yes, but it still needed to hack through the encryption, and it had no time to do this.

”Do you trust me?”

Here we go. Was I going to be gaslit like SecSys now? ”Yes,” I still said, because I did trust it.

”Then trust me when I say that Art is a friend.”

That wasn’t too bad. That didn’t feel like gaslighting even a little bit. ”Okay.”

”How are you feeling?”

”Okay.”

”No you’re not. How are you feeling?”

I was definitely getting the SecSys treatment now, yes. I could see why SecSys fell for it. It felt good, to be cared for. ”Performance reliability 89% and climbing,” I said, because I didn’t know how else to answer this question.

”Good.”

Then it got the Feed Monster to say out loud that it wouldn’t hurt me if I didn’t hurt it, its humans, or the Rogue first. This was reassuring to hear and I was grateful.

It appeared to me now that the Feed Monster had attacked me simply because it thought I was a hostile. If it was the Rogue’s friend, then of course an unexplained Combat SecUnit in its presence was a security concern. I couldn’t even blame it — I was also a shoot first, ask questions later type of bot. And it hadn’t actually killed me, even though it clearly could. I was starting to suspect this was only because the Rogue was shaming it every time it killed someone, like it had done with me.

I was starting to calm down. I was alive, the Rogue was alive, and the scary Feed Monster was kind of on our side. This wasn’t too bad.

Then the Feed Monster shouted at the Rogue for hiding that it was getting a constant 34% correction rate from its govmod. I also suddenly wanted to shout because what the hell, and why didn’t it tell me??

Because it still didn’t trust me, of course. And maybe it was right. I had even noticed that something was wrong, but I hadn’t asked. I should have asked.

I asked now. It tried to ignore me, and when I didn’t let it, it just blocked me.

I didn’t know what to do. I wanted to help, but I didn’t know how. I wanted it to let me look at its hack, because figuring out a patch would be faster together. But it didn’t trust me, and I didn’t know how to make it to trust me.

(I was so compromised. I didn’t even care anymore. The Rogue freed me. It got me out. I think I was allowed to be compromised about this.)

The Feed Monster apparently had a similar idea. I was starting to reluctantly like it. When it demanded the Rogue to share what it knew about the govmod and it refused, I sent the Feed Monster the specs myself, because I had them too.

The Rogue glared at me in the feed. I had seen it glare better than that. It also started to leak its emotional metadata into the feed again, and I could see it wasn’t actually angry. There were a lot of of things in there and I couldn’t fully parse them, but it appeared that the Rogue was very happy and very confused about this.

I thought its organics were just making it stupid again, like back then with the comms device.

The Rogue had eventually calmed down and settled back into its chair, and it didn’t even look as uncomfortable when doing this anymore. I could see the Feed Monster wrapping itself around it in the feed, and based on the emotional metadata, this was making the Rogue feel better. Because it was crazy, apparently.

I slid all the way down the wall to sit on the floor, and just stayed there. The Feed Monster unexpectedly pinged me for my diagnostics, politely staying as far away from my walls as a giant feed monster hypothetically could.

I sent the diagnostics. It felt like a polite thing to do.

The Feed Monster took barely any time at all to process them. You’ll be okay, it said.

That sounded unrealistic. But I thought it might actually be right.

I would be okay.