Valve finally brought Overwatch back to Counter-Strike earlier this week with its new CS2 Video Review tool, but skeptics are already pointing to a critical flaw: cheaters appear to be hijacking the system by inviting VAC-banned accounts into the program. Reports circulating on social media and within the community show that the invite-only VACNet Labeling Portal can be accessed by accounts with active VAC bans, raising immediate concerns about whether the system can be trusted to train Valve’s anti-cheat AI. Yet the reality is more nuanced—and Valve may already have measures in place to neutralize bad actors before their verdicts ever influence the model.
Unlike the original Overwatch system that debuted in CS:GO, the new browser-based CS2 Video Review is far more condensed. Reviewers are shown only a 10- to 15-second replay clip—not highlights from an entire match—and asked to provide a verdict. Available labels include Aim Assist, Wall Hack, Auto BHop, and Bot Player, along with an “Uncertain” option, which is explicitly encouraged when the evidence is ambiguous. The entire process happens outside the game client, through the VACNet Labeling Portal, and participation is by invitation only. Either Valve issues an invite directly, or a current reviewer with a proven track record of accurate verdicts can extend an invitation to others. The exact criteria for gaining invitation privileges remain unpublished, but the community has already discovered a loophole: users report that VAC-banned accounts can be invited to the program.
How the New CS2 Video Review System Works and Why Its Design Matters
To understand the potential for abuse, it’s essential to grasp what the new system is—and what it is not. Valve has described CS2 Video Review as a data-gathering mechanism for its VACNet machine learning models. Unlike the original Overwatch, where human verdicts could directly result in a temporary or permanent ban, the new tool uses reviewer decisions solely to train the AI. This shift fundamentally changes the stakes of any single verdict. A mistaken—or malicious—label does not immediately penalize the suspected player. Instead, it feeds into a dataset that Valve uses to improve automated detection over time.
The invitation-only design is intended to limit participation to reliable reviewers. By restricting access to trusted individuals and allowing vetted users to nominate others, Valve hopes to build a curated pool of human labelers. In practice, however, the system’s reliance on a chain of invitations creates a vector for exploitation. If a reviewer with invitation rights is compromised—or simply acting in bad faith—they can extend access to others who share their intentions. And when those new invitees happen to be VAC-banned accounts, the integrity of the entire labeling pipeline comes into question.
Reports of Cheaters Inviting VAC-Banned Accounts to the VACNet Portal
On August 13, 2026, Twitter user Ozzny posted a video clip showing cheaters blatantly misusing the new system. In the clip, reviewers are seen labeling obvious cheating clips—likely including aim assistance and wall hacks—as “not cheating.” Ozzny’s post explicitly states that cheaters are inviting other cheaters in order to inflate the number of fake labels, thereby corrupting the training data for VACNet. Another user, going by the handle poggu__, reported that VAC-banned accounts can indeed be invited to the program, a claim that quickly spread across Counter-Strike forums and social media.
“Cheaters are already trying to abuse CS2’s new Overwatch system. As shown in the clip, they are labeling cheating clips as ‘not cheating’. It’s also reported that cheaters are inviting other cheaters in order to increase the number of fake labels.” — Ozzny (@Ozzny_CS2), August 13, 2026
blockquoteblockquote
The evidence, while anecdotal, aligns with long-standing fears in the community: any human-in-the-loop system is vulnerable to organized sabotage. The previous Overwatch in CS:GO faced similar issues, though its direct-ban mechanism meant that malicious verdicts could harm innocent players. The new system, by decoupling verdicts from immediate punishment, changes the calculus. But it also introduces a subtler threat: if VACNet is trained on tainted data, its future detection accuracy could be compromised, potentially allowing cheaters to fly under the radar for months or longer.
Why Valve Likely Has Countermeasures—and Why They Might Not Be Enough
The immediate counterpoint to these concerns is that Valve is almost certainly aware of the potential for abuse and has built in safeguards. The most straightforward countermeasure is account correlation: Valve can track which accounts are VAC-banned and silently disregard any verdicts submitted by them, as well as flag the accounts that issued the invitations. In this scenario, the contributions from banned accounts would be weighted at zero or simply excluded from the training dataset. Valve could also assign lower confidence scores to verdicts from accounts with suspicious histories, or require a minimum number of consistent, accurate labels before a reviewer’s ratings are considered reliable.
Further evidence of a layered system comes from the VACNet portal’s design. The “Uncertain” option exists explicitly to reduce noise—reviewers are discouraged from guessing. Valve has also indicated that professional Counter-Strike players and other highly trusted individuals may have their verdicts weighted more heavily. If such a tiered voting system is in place, a small cabal of cheaters flooding the portal with false labels would have minimal impact on the final training signal. Theoretically, Valve could even run internal validation checks by injecting known-positive and known-negative clips into the review queue and comparing the reviewer’s answers to the ground truth.
Nevertheless, the success of these countermeasures depends on Valve’s ability to detect and isolate bad actors quickly. If cheaters can generate a large volume of false labels before being identified and removed, some tainted data could slip into the training set. Moreover, the opacity of the system means the community has no way to verify whether Valve is actually filtering out VAC-banned accounts or relying on reputation scores. The company has made no public statement on how it handles invitations from banned users, leaving players to speculate.
Key Differences from the Original Overwatch in CS:GO
The transition from CS:GO to Counter-Strike 2 in 2023 left Overwatch entirely absent from the game. For over two years, the community had no official human review system for cheating reports, relying solely on VAC and VACNet’s automated detection. The return of a review tool was widely welcomed, but the changes between the old and new systems are significant.
| Feature | Original Overwatch (CS:GO) | CS2 Video Review (VACNet Portal) |
| Verdict impact | Direct bans or timeouts | AI training data only |
| Replay length | Multiple highlights from a match | 10-15 second single clip |
| Access method | In-game application based on ranking and stats | Browser-based, invite-only |
| Reviewer pool | Open to high-rank players meeting criteria | Curated by Valve and existing reviewers |
The condensed clip format is designed to reduce the time burden on reviewers and speed up data collection. However, it also makes it harder to judge context—a single 15-second clip may not reveal subtle wall-hacking that only becomes apparent over a longer sequence. Valve’s response is that uncertainty is acceptable, and reviewers should use the “Uncertain” label when in doubt. The downside is that cheaters exploiting the system need only label a handful of short clips incorrectly to pollute the dataset, whereas previously they would have had to review entire matches to influence a ban.
What This Means for the Future of Anti-Cheat in Counter-Strike 2
The controversy over VAC-banned accounts being invited to the VACNet portal underscores a broader tension in anti-cheat design: the trade-off between human oversight and scalability. Automated systems like VACNet can process millions of reports, but they require high-quality labeled data to improve. Human review can provide that quality, but it is slow, expensive, and vulnerable to manipulation by determined actors. Valve’s hybrid approach—using a small, invite-only human panel to train an AI—is a logical middle ground, but only if the panel can be kept clean.
Long-term, the success of CS2 Video Review will depend on Valve’s ability to rapidly iterate on its filtering mechanisms. If the company can demonstrate that malicious verdicts are being detected and discarded, community trust may grow. If not, the system could become a net negative, providing cover for cheaters who can point to flawed training data as an excuse for missed detections. The situation also highlights the need for transparency: a public dashboard showing how many accounts have been invited, how many verdicts have been submitted, and what fraction came from flagged accounts would go a long way toward reassuring skeptical players.
For now, the most important takeaway is that the new Overwatch system does not directly ban anyone. Even if cheaters manage to submit a thousand false “not cheating” labels, those verdicts only become harmful if VACNet learns the wrong pattern. Valve can reset or reweight the training data at any time, and it likely maintains a gold-standard test set to validate model performance. The real question is whether the company will act quickly enough to prevent the dataset from being polluted in the first place, and whether it will communicate those actions to a community that has grown weary of promises about better anti-cheat.
Father time will tell. Valve’s track record with anti-cheat in Counter-Strike 2 has been mixed—VACLive improved matchmaking bans but did little to stop the proliferation of cheat providers. The reintroduction of a human review layer is a step forward, but the very fact that VAC-banned accounts can slip through the invitation process suggests that the implementation is still rough. As the system matures, the community will be watching closely: not just to see if cheaters can abuse it, but whether Valve can outpace them.
