How Rainbow Six® Siege Took Back Its Voice Chat with ToxMod to See a 50% Drop in Critical Toxic Voice

Abstract

Ubisoft’s Rainbow Six^® Siege (Siege) is one of the most competitive and popular first-person shooters in the world, reaching over 100 million players worldwide since it launched in December 2015. Within Siege, voice chat plays a significant role in how players experience the tense, second-by-second action, and communicate with their teammates. It’s a core strategic aspect of the game. For nearly a decade, however, voice chat was a challenge for Ubisoft’s moderation teams.

This case study dives into how Ubisoft’s Rainbow Six Siege team deployed Modulate’s ToxMod to gain more visibility into voice, resulting in per-session toxicity rates that fell from roughly 37% to 8% within two weeks going live, and a 50% reduction in critical toxic voice chat volume.

Beyond the harm reduction, ToxMod gave Ubisoft access to reliable, defensible data. Patterns of hate speech, repeat-offender behavior, and enforcement opportunities that were previously difficult to assess are now shaping how the Siege team better protects and builds its passionate community of gamers, and how it thinks about rewarding positive play heading into the future.

Executive Summary

In 2025, Ubisoft made Rainbow Six Siege a free access title. The anticipated wave of new players garnered excitement among the team, but the switch to free access also opened the door to an influx of potentially toxic behavior no longer gated by a premium price tag. At the time, the team was looking for a way to better review voice chat or moderate content at-scale.

Looking for the right solution, the Siege team kicked off a competitive evaluation process and a four-week trial with Modulate’s ToxMod to characterize the in-game environment and tune detection thresholds for their voice chat content moderation needs. Post-trial, Ubisoft and the Siege team decided ToxMod was the best solution to solve their challenges and went live in production in September of 2025. The impact was immediate and measurable.

The data tells a story beyond detection alone. In the weeks following go-live, both the toxicity rate and absolute incident volume declined quickly, meaning there was strong evidence in the data that player behavior actually changed. The additional visibility also provided intelligence that wasn’t possible before: ToxMod surfaced 9.5 million high-severity incidents across the first eight months, revealing that on any given day, roughly 35% of flagged users drive 66% of all incidents — insights that now help shape Ubisoft’s overall enforcement strategy.

Background

The Most Competitive Ubisoft Title (and hardest to moderate)

Siege occupies a unique position at Ubisoft: it’s the company’s flagship online competitive first-person shooter, a ten-year-old title, and home to a fiercely dedicated player base. And by design, it’s intense. When you die in Siege, you don’t respawn like in other games. You watch from the sidelines. Friendly fire and team kills are possible, every round is high-stakes, and some amount of frustration is baked into the risk/reward experience.

Voice chat sits at the center of all of it. Siege provides open-mic team voice for every player in every match. It’s not proximity-based chat that fades in and out like other shooters based on how close you are to teammates. If you’re using voice chat in Siege, you likely know the game well: the callouts, the maps, the meta progression of weapons and operators. Communication isn’t simply a social garnish. It’s a way to create a competitive advantage.

The combination of a high-pressure competitive format plus always-on, full-team, voice chat makes the likelihood of toxicity meaningfully higher in Siege than in most other noncompetitive games. The Siege team knew this was a factor to consider, however, they needed a way to see it.

The Problem

A Decade of Voice Chat Without Sufficient Visibility

Lancelot Satge, Rainbow Six Siege’s Game Protection Director, said this about how the team handled toxic voice behavior before ToxMod:

Text chat could be filtered and reports could be filed. But voice chat was more complex. Ubisoft’s broader approach relied on manual processes distributed across multiple teams, which preserved human judgment but couldn’t deliver the responsiveness a live-service game demands on a continuous basis. Voice made it harder because it’s inherently contextual, nuanced, and impossible to interpret at scale without dedicated tooling like ToxMod.

For the gaming industry as a whole, it’s no secret that toxic interactions can degrade the quality of play and impact player wellbeing. For a game as popular as Siege, this reality was unavoidable and something the team took seriously. In Ubisoft’s weekly community sentiment reports, which screen social media for the most popular player topics, voice chat toxicity appeared in 2–4% of all discussions for five years straight.

By 2024, growing report volumes and limited visibility into voice interactions had already made automated voice moderation a strategic priority. The decision to open up Rainbow Six Siege as a free-to-access title further reinforced that priority. With the community expected to grow, Ubisoft and the Siege team focused on ensuring that player safeguards could scale with it from day one.

‍

The Solution

Enter ToxMod, The Most Complete Solution for Siege

Ubisoft ran a formal Request for Proposal process with cross-functional stakeholders to find the best solution. The evaluation criteria were clear: voice model precision, operational flexibility, and effectiveness compared to taking speech-to-text data and running it through a text moderation tool, or a traditional LLM. ToxMod demonstrated stronger detection quality at a comparable cost, but with significantly better granularity than the alternatives evaluated. The deciding factor was simpler according to Lancelot:

Ubisoft’s Global Player Safety team had first encountered Modulate at Trust & Safety industry events and through early deployments in immersive and VR environments where voice is central. What stood out was ToxMod’s maturity: a solution already battle-tested in real production contexts.

A four-week trial built the confidence to go live

Rather than flipping the switch, the partnership began with a four-week non-enforced trial. ToxMod ran live against real Siege audio, characterizing the in-game environment and tuning model thresholds while no player-facing action was taken. This phase surfaced one of Siege’s trickiest moderation challenges due to the specific vocabulary players used in a game about counter-terrorism operations. For example, “let’s kill him” is a callout in Siege, not a threat. The system had to learn the difference between in-game vocabulary and genuine harm. No easy task to say the least.

Key findings: what visibility revealed

Once live, ToxMod began processing ~800,000 hours of Rainbow Six Siege voice audio per month. What it revealed was incredibly telling:

• • The sharpest signal was also the hardest to catch. The behaviors least visible in live voice were the ones the system could surface most reliably. Additional visibility gave the Siege team something it hadn't had before with a high-confidence stream to tune thresholds against and route to enforcement, instead of a backlog of reports to triage.
  • A small minority drives the majority of harm. On any given day, about 35% of flagged users generated two or more high-severity detections and those users accounted for 66% of all incidents. The pattern held on every single day of the analysis window.
  • Protection goes beyond toxicity. Ubisoft also enabled ToxMod’s user behavior tracking which flags for continuous harmful patterns that represent real risk to players. The behavior tracking satisfied Ubisoft’s strict global safety and reporting requirements.
    ‍

The Results

ToxMod Impact: Visibility Improves, Toxicity Decreases

Enforcement went live on September 17, 2025. Day one, ToxMod flagged 58,172 incidents across 21,234 unique players, a 2.5x surge over the prior week’s baseline as the system began processing data. Then something better happened: players adjusted.

Within two weeks of in-game warnings and mutes going live, the per-session toxicity rate dropped from roughly 37% to 8%. The rate and the absolute incident volume fell together, which the data identifies as the clearest evidence of behavioral deterrence, i.e. players weren’t just being caught, they were considering their individual behavior. Toxic recordings also dropped by 50% in volume practically overnight.

The community’s response followed a pattern Ubisoft’s safety team has seen before with some players perceiving stronger enforcement as censorship, while others self-adjusted as the community recognized the benefits of a more positive environment. The net result? A clear positive impact on overall trust.

Internally, the data did something no policy document ever could: it ended the debate about whether the problem was real.

From visibility to enforceability

An enforceable code of conduct requires evidence and, for voice, Ubisoft finally has it. The repeat-offender insight is now shaping a graduated penalty framework that escalates consequences for repeat toxic behavior, applied with the precision to avoid punishing a player for what may appear as a heated argument but, in reality, is harmless banter between friends. And because the most toxic players are often the most invested players with hundreds or thousands of hours in game, deep game knowledge, and a competitive nature, the team can now reach for smarter tools than the “ban hammer,” like removing a repeat offender’s right to speak in-game while letting them continue to play.

Following the initial ToxMod implementation, internal teams pivoted together to a “targeted players” model to focus ToxMod’s analysis on users already reported or moderated above a set threshold, delivering a highly accurate, immediately actionable stream to Siege’s enforcement teams. It’s now a matter of continuous improvement and fine tuning thresholds as needed.

Now after a full year, Ubisoft’s Siege team renewed its partnership with Modulate, evolving the deployment toward a more player-driven model. As Lancelot frames it, the first year was about giving the team “the freedom to learn” how players react, what enforcement takes, and what a healthy Siege sounds like.

What’s Next for Positive Play

The Rainbow Six Siege story is proof of something the industry can learn a lot from: in games where voice is core to play, voice safety is key to a positive player experience. Siege didn’t shy away from the challenge of voice moderation at scale, or treat it as a compliance checkbox. It treated toxic voice as a serious issue to address, and the results show up everywhere from session-level toxicity rates to the community’s weekly sentiment reports.

For Lancelot and team, reduced toxicity is the foundation for continued improvements. He sees an opportunity to not only detect toxicity, but to detect positive players and find ways to reward or thank them. These would be the players giving great callouts to their team, leading by example, or welcoming newcomers. With ideas like these, the future of Siege, and gaming as a whole, is bright.

Advice for teams facing a similar situation

For gaming studios facing the a similar problem, Lancelot’s advice is spot on:

• Know why you have voice chat. Voice carries real costs: moderation, tooling, and balancing. Understand the ROI before you commit.
• Understand your specific game’s reality. Open-mic team voice in Siege is not proximity chat in other competitive shooters. Your format shapes your risk. Understand your game, what makes it unique, and how voice chat fits in.
• Know your community and your vocabulary. A competitive community tolerates different things than a community of younger players, and your in-game language (operator names, objective callouts) must be built into moderation from day one.
• Don’t forget positivity. Fairness and player-driven solutions are the key and the end goal isn’t silence, it’s better communication.
  ‍

Rainbow Six Siege, one of the best and most popular competitive shooters in the gaming industry, spent ten years without proper visibility. With ToxMod, they heard all of it and were able to better enforce trust and safety standards. The proof isn’t just in the 50% reduction in critical voice chat events, the 75% drop in per-session toxicity rate, or the five-year community-sentiment signal that vanished in months. It’s in the fact that players have a more positive space to enjoy something they hold near and dear to their hearts. That’s what winning looks like in trust and safety.

For Siege, the community is worth protecting for another decade. For Modulate, it's proof that even the most competitive voice chat can be a positive space for players with the help of a solution like ToxMod. Heading into the future, the best rounds are yet to come.