Improving Anti-Cheat: Server-Side AI Detection (Vacuum, Speedhack, Lag Switch)

Conan Exiles General Discussion Feedback
1

Hello,

I would like to suggest an idea for improving protection on Conan Exiles official servers. Currently, BattlEye does a good job catching external programs, but some exploits remain undetected, such as:

  • Vacuum (accessing containers from a distance)

  • Speedhack (abnormally high movement speed)

  • Lag switch (artificial network delays to gain advantage)

Since these exploits rely on server-side logic, they can only be reliably detected at the server level. One possible solution would be an AI-based system — a sort of “AI-admin” — that monitors anomalous player behavior through simple checks (movement speed limits, object interaction ranges, network lag patterns). This would act as a strong complement to BattlEye, enhancing its overall effectiveness rather than replacing it.

How this would reduce support workload:

  • Fewer manual reports, as suspicious actions are automatically flagged.

  • Automatic warnings, shadowbans, or temporary kicks allow responses without human intervention.

  • Manual review is needed only for borderline cases, saving the support team’s time and resources.

This approach could significantly reduce cheating and improve the gameplay experience on official servers. I would be very interested to hear the developers’ thoughts on this idea.

  • :+1: Yes — I support the idea of an AI-admin; it would help reduce cheating and improve gameplay
  • :thinking: Maybe — I like the idea but would like more details on how it works
  • :-1: No — I don’t think this is a good solution or necessary
0 voters
1 Like
2

Phased implementation:

Start with simpler cases like vacuum exploits and unauthorized bases (flying bases, undermesh), which are easier to detect and already cover a large part of cheating issues. Gradually add more complex detections like speedhacks and lag switches. This phased approach reduces risk, shows results faster to players, and makes it easier to expand the AI-admin system.

3

So what you’re saying is you’ll be perfectly happy when you get banned because of an AI “hallucination”, when the AI spontaneously makes up fake data about your account that no one can ever disprove.

2 Likes
4

The AI-admin is not meant to ban players automatically based on a single event. Instead, it acts as a support tool for the server and developers:

  • Suspicious activity is flagged, not immediately punished.

  • Multiple indicators are considered (movement patterns, object interactions, network behavior, reports from other players).

  • Players are usually given warnings, temporary kicks, or shadowbans, with manual review for borderline cases.

  • Feedback from affected players can be used as an additional factor to reduce false positives.

:white_check_mark: The goal is to assist human moderators, not replace them, and to minimize false bans while effectively detecting cheating.

5

PS. The AI cannot “make up” data if it is only monitoring server logs and events. It simply analyzes the existing records. The AI reads these logs and looks for anomalies — it does not modify or “create” events.

1 Like
6

Google Gemini can’t even give reliable tips on how to use configure certain settings in something like Google Sheets.

And while you say that the goal is to assist human moderators, likely to help them spot these anomalies and then direct their investigations to confirm/refute, I can say with 100% certainty that at least some people will assume the AI is correct and action based on those results without digging deeper.

7

The AI-admin collects information only from existing server logs, which can always be checked. Its role is to filter and structure the data, so moderators don’t have to manually review huge amounts of information. They receive summaries and highlights of the areas that need attention, which speeds up investigations and reduces workload.

In short, the AI does not create new data, it helps moderators use existing information more efficiently and faster.

8

Honestly, this reads more like an AI ad than an actual attempt at conversation around the well-known issues currently surrounding implementation of AI tools.

3 Likes
9

The goal is not to advertise AI, but to discuss practical ways to help moderators and players, since cheating is currently a real problem. The main focus is on improving moderation efficiency and protecting fair play, not promoting AI.

And by the way, returning to your previous comment. If there are complete server logs, there’s no need to dig much, because they already contain all the key information.

10

There’s 2 problems with this.

  1. You can ditch the AI.. there is absolutely no need for it, so in this whole idea pitch that seems to revolve around AI, the AI is actually the part that’s pointless :slight_smile: This stuff can be detected with very simple purpose coded checks.. somehow involving AI would just make it far more complicated and a waste of resources.
  2. Good luck implementing these on the 5 dollar gportal servers that officials are hosted on..

(Edit: just so I’m not leaving without suggesting alternatives: the way this should be working ideally is that the developers come up with workarounds for these hacks and exploits and implement security measures directly into the game, so that it applies to all servers and the cheats just stop working on the game.
Funcom has actually been doing this quite often in the past, but the thing is.. it’s a cat and mouse game and there are people actively developing cheats with at least that much effort or more. It doesn’t help that lately Funcom shifted focus to Dune, so they will be putting way less effort into any of this whether it’s AI or otherwise than they did before.)

11

If everything is supposedly so simple, why are reports still being reviewed over weeks then?

12

I completely agree that the best solution is game-level fixes that stop exploits for all servers at once. The AI idea is meant as a temporary or supportive tool that can help detect and respond to cheaters faster, especially when developers cannot immediately fix all exploits.

:white_check_mark: The goal of this idea is to reduce moderator workload, protect players, and speed up cheat detection, not to unnecessarily complicate the system.

The most important point is that this is an endless cat-and-mouse game: developers fix exploits, and cheaters find new workarounds. That’s why quick detection and banning of cheaters is also highly effective. There’s no point in cheating if you get caught and punished within short time.

13

I’m not sure you understand it.. just because AI is the cool new toy, does not mean it has to be shoved into everything.. or that it would even be possible or reasonable to do so.

The AI we have is good at certain things.. the popular versions are good at searching the internet and trying to formulate responses based on that or make up “art” by copying actual artists and mixing and matching them. It’s basically a bulk operation, or doing a lot of things in a short time.
That’s where it’s useful and helpful.. because it can skim through way more articles per second than you can.. and you still need to have some form of common sense, because it will lie to you with a complete poker face.
Sure there are other applications too, but generally speaking this is where it shines.. being able to process large volumes of data and then trying to in some way communicate that to someone or “get an idea” of what it just processed.

Where do you see any of that being useful in this case?
Let’s go with some of the examples from your post:
Vacuum:

  • Items are being transferred from one inventory to another..
  • The player isn’t within interaction range of the object that houses the source inventory.

Do we need any more information than that?.. Does the internet need to be searched for this? Nope.. It’s basically
 “The player is remotely looting stuff”. How would AI be of any help here?.. is there a large volume of data? Nope.. this is something very simple code can straight up prevent. Does this need to be summarized and communicated to anyone?.. nope..

So what does AI can do that somehow goes beyond a developer adding this to the game code :

image
image1630×623 144 KB

(ofc expanding this simplified version a bit to account for harvesting and mods with pull from nearby inventory features etc. possibly putting it behind an official server setting)

This took less than a minute and essentially nullifies that exploit forever
 It would be a complete waste of processing power not to mention the ridiculous amount of work (months) required to somehow tie a 8 year old game into some weird AI admin.

There is simply no need for it
 what is needed is the devlopers monitoring active cheats.. and implementing security features like the above to make them not work. But in any case it requires more resources than what is currently allocated to this game.

Official server pvp is just not the prime target audience of this game so what little dev time they allocate to this project is going to be mostly spent on fixing actual game bugs that happen constantly without a player going out of their way to break the system with external programs. We still have a lot of those around.

There are NO active moderators
 people send reports and some of the Funcom customer support staff checks it out occasionally.

Because there is barely anyone left working on this game. Supposedly there is a “team” but we don’t know who or how many. What we do know is that they have HALF a community manager left on this game (Andy), who is also assigned to Dune as far as I know, but occasionally posts some Conan stuff.

And any kind of work, whether it’s AI or not.. does require people to do it.

3 Likes
14

The key difference here is that the game code (the client side) is vulnerable to cheaters — they can modify it, bypass checks, or manipulate packets. That’s why simple checks built into the game itself will eventually be circumvented.

The server side, however, is closed to players. A cheater cannot alter server logs or the server’s validation logic. This is why server-side analysis and automation with an AI-admin is much more reliable: it works with data that cannot be faked from the client.

15

I’m not sure what you mean by “key difference”
 the above code I posted actually runs server-side..
As in the image I posted there I did literally in the game code of Conan Exiles via the devkit and it’s not some arbitrary example.. if I were to build that as a mod, it would work straight up, but ofc Funcom needs to be the one to implement a similar solution and update the game with it for everyone.

1 Like
16

I don’t intend for this to sound sarcastic when I say, good luck with that. Just about every company on the planet is trying to figure out how to use AI instead of people, even when they say otherwise.

Game companies, in particular, are notorious for trying to find the easiest and cheapest methods they possibly can to administer their game environments. Gaming is a retail business with essentially zero consumer protection. Think about how many people have been falsely banned by third party tools like BattlEye, then multiply that by the speed and volume of AI.

I work with AI every day, and it’s only effective with lots of human oversight. Any game company that is considering AI for moderation is already aiming for the target of zero human oversight, or as little as they can possibly get away with.

While I admire your optimism (truly I do, again that’s not sarcasm) you have to look at how FC is already doing every thing they can to spend as little money on Conan as they can. The reasons thet keep this game alive are:
a) So no one can claim it’s abandonware.
b) So they can brag bout it as part of their portfolio, “Look at all the cools games we created, still own, and ‘support’ (even though some of those games are no longer profitable)”.

Maybe, just maybe, they would learn how to create AI agents to manage cheating in Dune, where they’re still making enough money to make it worthwhile, and then after the AI agents are already doing a good job FC might apply them to Conan. If that were to happen it might actually have benefits for Conan. But even that is a highly dubious proposition, in part for reason that Xevyr has already explained.

And if they attempted to us AI moderation in Conan first, where they would do it with no budget, it would be the worst version of AI moderation people can imagine.

3 Likes
17

Yes, it absolutely can. All of them do it and they do it every day. The AI companies use the euphemism “hallucinations” but what they really mean is their AI just makes up garbage out of nothing and they don’t know why.

There are tons of articles and videos explaining how this works. Just google “AI hallucination” and you’ll see that it happens all the time.

18

I understand that this is your goal. However I can tell you from first-hand experience that even when you’re using AI in exactly the way you’re describing, it can still “hallucinate” data that doesn’t exist.

19

A number of lawyers and expert witnesses have been sanctioned by courts for citing made-up, or “hallucinated,” cases and facts generated by artificial intelligence tools like ChatGPT.

Ironically that sentence was generated by ai when I did a search for the court cases I’d read about previously. It’s very bad. So yeah, trust AI about as much as three day old potato salad.

20

You’re right that in the case of simple cheats like vacuum, a straightforward server-side condition, like in the blueprint example you showed, solves the problem permanently — and there’s really no need for AI in that case.

But there are more complex situations, like speedhacks or lag-switching. In those cases, simple checks aren’t enough. That’s where a tool like an “AI-admin,” which analyzes logs and detects patterns, could actually be useful.

I can already guess what you might say in response — something along the lines of: “If Funcom can’t even implement simple fixes, how could anyone expect them to handle something as complex as AI?” And unfortunately, I would have to agree. If basic issues have remained unaddressed for years, it’s hard to expect more advanced solutions.

That said, I hope Funcom will finally pay attention to this problem — especially since they themselves stated that they intend to improve the existing game. Ignoring cheating is exactly what drives players away, and it damages not only Conan but also Dune.

And for me personally, this has already had a direct effect: I changed my mind about buying Dune. There are already plenty of videos showing cheaters there too, and if Funcom doesn’t address this, people will simply stop trusting their games.