Roblox CorpRBLX

Roblox is shipping professional AI agents including a Playtesting Agent to creators.

“​impressive. You can see several examples of live​​gameplay​​from Build - we can't wait for what’s to​ ​come.​ ​AI Momentum and Adoption.​​Creator use of AI tools​​is scaling rapidly, with adoption rates among​ ​our top 1,000 and top 10,000 creators by Robux spent up approximately 15 percentage points​ ​quarter over quarter. Enhancements to our AI-native Studio helped drive a 20% sequential​ ​increase in creators using our agentic Studio Assistant. Additionally, in a survey of several​ ​thousand Roblox creators who use AI for development work, 95% said AI speeds up their timeline​ ​to launch new games or game updates into production, of which nearly half experienced an over​ ​50% acceleration in publishing timeline. These advancements reinforce our vision to redefine the​ ​future of Play—freeing creators from tedious technical work so they can focus on high-value​ ​creative design. Adoption of our 3D generative AI tools continues to scale. Creators are now​ ​generating over 60,000 3D assets per day and nearly 1,400 games use Cube-generated content​ ​daily—early evidence that these tools are gaining traction with creators ranging from emerging​ ​developers to some of the platform's largest experiences.​ ​Advanced Creator Agents.​​Over the next several months,​​we are shipping a suite of professional​ ​AI agents across Build and Studio to further streamline development: 1) Playtesting Agent​ ​automates QA and bug detection, eliminating the need for manual testing, 2) Analytics ”

Roblox notes threat actors increasingly use agentic AI for sophisticated cyberattacks.

“The techniques used by malicious actors to obtain unauthorized access to, or to sabotage, our systems or networks, or to utilize our systems maliciously, are constantly evolving and generally are not identified as a threat vector until launched against a target. Threat actors are also increasingly leveraging AI, including agentic AI systems, to conduct reconnaissance, exploit vulnerabilities, craft highly personalized social engineering attacks, and adapt their methods in real time without human intervention. These AI-driven techniques may enable attackers to operate at a speed, scale, and degree of sophistication that outpaces our ability to detect and respond to threats. In addition to actions by individuals, such attacks could also potentially be launched by nation states, state-sponsored bad actors, or other well-funded and highly sophisticated individuals. Despite the measures we have taken, we at times have not been able to anticipate these techniques by implementing preventive measures, detecting, or reacting in a timely manner, which has resulted in and could continue to result in, delays in our detection or remediation of, or other responses to, security breaches and other security-related incidents. In addition, the use of open source and third-party software in our Platform has exposed us to security vulnerabilities in the past and will likely continue to expose us to security vulnerabilities in the future.”

Roblox warns that deploying agentic AI adds cybersecurity risks like prompt injection and data exposure.

“The use of AI in our products and business practices may increase or create additional cybersecurity and privacy risks, including risks of data breaches and security incidents. In particular, the deployment of agentic AI introduces new attack surfaces and risk profiles for us to address. Prompt injection attacks, in which malicious instructions cause an AI agent to take unintended or unauthorized actions, are an emerging threat vector for which established defenses are still developing. As AI use expands across our products and operations, we expect the frequency and sophistication of AI-targeted attacks to increase. Beyond external threats, agentic AI systems operating with limited human oversight at each step may also inadvertently expose proprietary or confidential information by autonomously executing tasks across data boundaries without explicit human authorization.”

Roblox incorporates and deploys AI in its platform, raising risks of flawed outputs and manipulated content.

“We have incorporated, and are continuing to develop and deploy, AI in our products and the operations of our business. Our use of AI in aspects of our Platform may present risks and challenges that could increase as AI solutions become more prevalent. The Roblox Cloud may be more relied upon in the future to facilitate increasingly complex decision-making as it integrates hardware and accelerated machine learning and other AI, for a broad range of compute tasks, including improved personalization, synthetic content generation, enhanced automation of the player experience, and AI systems deployed through our Platform. However, AI algorithms may be flawed and datasets may be insufficient or contain biased information. Even with safeguards in place and oversight, AI systems may make decisions unpredictably or autonomously, such as generating incorrect, offensive, and/or infringing content and may take actions beyond their intended scope. This can raise new or exacerbate existing ethical, technological, legal, and other challenges, and may negatively affect the performance or the perception of our Platform and the user and creator experience. Certain users have and may in the future attempt to manipulate AI systems to create violative content on our Platform. In addition, threat actors have used and in the future may use AI to enhance the effectiveness of their attacks against our Platform which may cause the loss of availability of the Platform, degradation in services, or compr”

Roblox discusses new AI regulations including the EU AI Act that may affect its AI use.

“There are also many new and evolving laws and regulations focused on the use of AI, and agentic AI systems that operate with significant autonomy may face heightened scrutiny or additional regulatory requirements. For example, the EU’s Artificial Intelligence Act (“AI Act”) entered into force in August 2024. Certain of its obligations entered into effect in February 2025, and many of its applicable provisions are due to become effective by August 2026, although the EU’s legislature may vary the dates for certain obligations, and the European Commission has published the Digital Omnibus, a legislative package that proposes to, among other things, delay certain high-risk AI system deadlines with transparency obligations under the AI Act largely proceeding on the August 2026 schedule. EU institutions reached provisional agreement on the Digital Omnibus in May 2026, although it remains subject to formal adoption. The AI Act establishes a framework of prohibitions as well as disclosure, transparency, and other regulatory obligations based on various levels of risk for businesses introducing AI systems in the EU, and the Digital Omnibus proposes to expand the scope of prohibited AI practices under the AI Act. Provisions of the AI Act could require us to alter or restrict our use of AI both in features or products available to our users and in our systems that interact with our users, depending on respective levels of risk-categorization, types of systems, and manner of use, as set ”

Roblox states threat actors increasingly leverage agentic AI for sophisticated cybersecurity attacks.

“The techniques used by malicious actors to obtain unauthorized access to, or to sabotage, our systems or networks, or to utilize our systems maliciously, are constantly evolving and generally are not identified as a threat vector until launched against a target. Threat actors are also increasingly leveraging AI, including agentic AI systems, to conduct reconnaissance, exploit vulnerabilities, craft highly personalized social engineering attacks, and adapt their methods in real time without human intervention. These AI-driven techniques may enable attackers to operate at a speed, scale, and degree of sophistication that outpaces our ability to detect and respond to threats. In addition to actions by individuals, such attacks could also potentially be launched by nation states, state-sponsored bad actors, or other well-funded and highly sophisticated individuals. Despite the measures we have taken, we at times have not been able to anticipate these techniques by implementing preventive measures, detecting, or reacting in a timely manner, which has resulted in and could continue to result in, delays in our detection or remediation of, or other responses to, security breaches and other security-related incidents. In addition, the use of open source software in our Platform has exposed us to security vulnerabilities in the past and will likely continue to expose us to security vulnerabilities in the future.”

Roblox warns that deploying agentic AI introduces new cybersecurity risks including prompt injection attacks.

“The use of AI in our products and business practices may increase or create additional cybersecurity and privacy risks, including risks of data breaches and security incidents. In particular, the deployment of agentic AI introduces new attack surfaces and risk profiles for us to address. Prompt injection attacks, in which malicious instructions cause an AI agent to take unintended or unauthorized actions, are an emerging threat vector for which established defenses are still developing. As AI use expands across our products and operations, we expect the frequency and sophistication of AI-targeted attacks to increase.”

Roblox notes evolving AI regulations including the EU AI Act may require changes to its agentic AI use.

“There are also many new and evolving laws and regulations focused on the use of AI, and agentic AI systems that operate with significant autonomy may face heightened scrutiny or additional regulatory requirements. For example, the EU’s Artificial Intelligence Act (“AI Act”) entered into force in August 2024. Certain of its obligations entered into effect in February 2025, and many of its applicable provisions are currently due to become effective by August 2026, although the EU’s legislature may vary the dates for certain obligations, and the European Commission has published the Digital Omnibus, a legislative package that proposes to, among other things, delay certain deadlines until 2027 or 2028. The AI Act proposes a framework of prohibitions as well as disclosure, transparency, and other regulatory obligations based on various levels of risk for businesses introducing AI systems in the EU. Provisions of the AI Act could require us to alter or restrict our use of AI both in features or products available to our users and in our systems that interact with our users, depending on respective levels of risk-categorization, types of systems, and manner of use, as set forth in the AI Act. The AI Act also may require us to comply with monitoring and reporting requirements. As a result, we may need to devote substantial time and resources to continue to evaluate our obligations under the AI Act and to develop and execute a plan designed to promote compliance. Noncompliance with th”