FACEIT unveils its AI anti-cheat: catching cheaters by how they move the mouse, not what software they run
FACEIT has announced Human Input Detection, a new machine-learning layer within its anti-cheat that completely changes the question the system asks. Instead of looking at what program you have running, it analyzes how you play: whether your mouse movements, clicks, and aim corrections are physically consistent with a human being.
It goes live on August 5, coinciding with the start of Season 9, and will apply to CS2 matchmaking, the ESEA League, and official qualifiers within FACEIT's ecosystem.
The problem it's built to solve
To understand why this matters, you have to understand why the classic model has fallen short. Traditional anti-cheats work like antivirus software: they maintain a database of known threats and scan for matches. The problem is that today's most advanced cheats are designed precisely to leave no trace. AI-based systems read screen data through external hardware — typically DMA cards — and translate it into seemingly legitimate mouse movements. They don't inject code into the game or modify memory in recognizable ways. To a signature scanner, that computer looks clean.
FACEIT's own figures show the scale: in May 2026, AI and DMA cheats already accounted for 40% of all cheating bans on the platform.
Human Input Detection targets exactly that gap, and also the other end of the problem: brand-new tools no security researcher has catalogued yet. A signature system can't detect what it has never seen; a behavioral model can notice that something moves strangely.
How it works
The system is trained on input data from millions of matches. From that it builds a model of how a person actually moves, aims, and shoots, and flags deviations matching the telltale patterns of artificial assistance. And it keeps learning: every match played feeds the model.
Two design details are worth highlighting. First, it requires no download: it runs server-side, with no additional installation for the player and, according to FACEIT, no impact on FPS. Second, and more importantly, a detection does not mean an automatic ban. The company is explicit: the system acts as one more signal that its integrity systems weigh alongside the rest of the anti-cheat data, and multiple corroborating signals are required before a ban is confirmed. An isolated suspicious sequence won't get anyone kicked.
FACEIT has also stressed that this does not replace its kernel-level anti-cheat, but works alongside it. The two layers cover different parts of the same problem.
The inevitable comparison with VACnet
Human Input Detection's architecture becomes especially interesting compared to Valve's. VACnet operates server-side: it analyzes the reconstructed behavioral data the server generates after processing the player's inputs. FACEIT's layer acts at the earlier point in that chain, reading the client's raw input stream before it reaches that reconstruction.
That difference isn't theoretical, and this week showed exactly why. Valve's July 29 patch fixed a bug that allowed cheat clients to avoid being processed by VACnet — that is, to corrupt or bypass the data feed the system depends on. When your detection relies on data the cheater can manipulate, the analysis never happens at all.
What the community says (and what to watch)
The reaction has been largely positive, with plenty of posts hoping Valve takes note. But legitimate reservations have also surfaced and deserve airing.
The first is false positives, the specter haunting any behavioral detection system. The second, closely related, is what happens to unconventional playstyles: players with extreme sensitivity setups, atypical peripherals, or unusual movement habits could sit closer to the margin than they'd like. The third is privacy, though here the design works in FACEIT's favor: by analyzing in-game inputs rather than scanning the user's system, the new layer is less intrusive than the kernel anti-cheat already in place.
The requirement for multiple corroborating signals before a ban looks designed precisely to cushion those risks. Whether it works in practice remains to be seen.
A packed Season 9
The feature doesn't arrive alone. August 5 also brings the new ban policy FACEIT announced last week, with cheating penalties rising from two to five years and a notable easing in the moderation of minor toxicity. And it follows a recent crackdown on smurfing: on July 27 the platform sanctioned 12,478 accounts for multi-accounting and matchmaking abuse, with 13,644 total penalties that day, and restored around 20 million Elo to more than half a million affected players.
The contrast with Valve draws itself. While the CS2 community has spent months demanding answers — to the point that a player has taken to writing his own open-source anti-cheat — FACEIT keeps stacking announcements in the same direction. And although Valve has shown signs this week with its VAC Live push, it still hasn't publicly explained its strategy. That's probably the most visible difference between the two: not just what each one does, but what each one says.
Source: FACEIT official announcement, Ozzny (X), TechTimes.