Amazon’s vision is to build a world where consumers are confident that they won’t fall prey to bad actors who impersonate our brand or anyone else’s. As part of this commitment, Amazon has launched SENTRIX, a new artificial intelligence (AI) technology that enhances our ability to identify and remove malicious websites even faster.
SENTRIX accelerates our proactive protection capabilities, analysing more than 50,000 suspicious URLs weekly, helping us detect and respond to threats more quickly. It combines three key components: advanced monitoring that scans for potential threats, AI-powered risk assessment, and streamlined takedown processes. By investing significant resources in proactive controls like SENTRIX, Amazon has increased successful phishing URL takedowns by more than 10% in 2025, enhancing our ability to prevent impersonation scams from impacting consumers.
The Global Anti-Scam Alliance’s (GASA) global state of scams report 2025 reveals 70% of adults around the world experienced a scam in the past year. While scams are not a new problem, the opportunities for scammers to commit their crimes have increased as access to sophisticated technologies increases. Scammers evolve their tactics and use social engineering to exploit consumer trust. Traditional tactics such as phone calls and text messages remain common, and impersonation scams continue to be the most frequent form of fraud worldwide, according to cybersecurity firm HoxHunt.
Modern phishing attacks see bad actors employing advanced visual deception tactics alongside traditional text-based scams. For example, these fraudsters carefully replicate brand logos, copy professional website layouts, and mimic familiar design patterns to create fake websites that appear legitimate at first glance. This visual mimicry, combined with social engineering techniques, makes it harder for customers to distinguish genuine websites from fraudulent ones—requiring equally sophisticated AI-powered detection systems to identify and remove these threats. From fake retail sites created outside of our store to emails impersonating brands like Amazon, these external attacks make proactive protection more important than ever.
Innovating to stay ahead of bad actors
SENTRIX is an important part of Amazon’s strategy to leverage AI to protect consumers from scams, using a combination of search optimization, risk assessment, and multiple AI technologies to create layered protection for customers. While SENTRIX significantly enhances our detection capabilities, no system can catch every threat. Customers should continue to remain vigilant and report suspicious activity.
The technology works through three key components:
- Advanced data monitoring that continuously scans for potential threats. Utilising a third-party search API to monitor search engine results from Google, Bing, Yahoo, DuckDuckGo, and other sources, any URLs identified as potential risks are automatically sent to our AI-powered risk assessment, which is an LLM-powered phishing detection model.
- Sophisticated risk assessment. The advanced AI system analyses both textual and visual elements of URLs flagged by our data monitoring systems—including logos, page layouts, design patterns, and graphical elements—to detect sophisticated phishing attempts that use visual deception tactics. It then evaluates the flagged URL with contextual data and assigns it a risk level. SENTRIX also learns from past detections, continuously improving its accuracy and speed through an adaptive learning system that remembers effective detection strategies. Through SENTRIX’s automated detection capabilities, potential phishing websites are detected in less than 24 hours after they emerge.
- Streamlined takedown processes to quickly remove harmful sites. Once the risk level is identified, SENTRIX sends the URLs with the highest risk level to a trusted third-party verification service, where human review of flagged content is conducted before any takedown is actioned. The system retrieves and tracks these verification and takedown results to support continuous improvement.
“Protecting customers from scams is a global challenge that requires innovative solutions,” says Scott Knapp, vice president of worldwide buyer risk prevention at Amazon. “By combining multi-modal AI technology with the expertise from our global teams – from machine learning scientists to fraud investigators – we’re able to proactively identify and remove fraudulent websites at scale. This helps us reduce the likelihood of customers encountering these threats, rather than waiting for them to report problems.”
Advancing responsible AI through customer protection
While SENTRIX initially focused on the takedown of English and Japanese phishing websites, the tool has expanded to read multiple languages websites, addressing the growing prevalence of scams in these regions. According to GASA, in 2025, 46% of Germans reported being scammed in the last 12 months, with shopping scams being the most common. In Spain, the financial impact of scams was €7.4 billion in 2024, equivalent to 0.5% of its GDP.
Beyond its technical capabilities, SENTRIX reflects Amazon’s approach to AI development: technology designed with clear customer benefit and social responsibility in mind. As highlighted in recent discussions with European policymakers, we are committed to purpose-driven innovation.
The system demonstrates key responsible AI principles through transparent detection methods, fair and accountable processes including human audits to prevent false positives, and robust measures that protect customers across multiple languages and regions. SENTRIX focuses on positive social impact, mitigating potential risks, and addressing concerns related to AI’s broader societal implications—reflecting Amazon’s commitment to advancing responsible AI practices across the industry.
Looking ahead
SENTRIX is just one example of Amazon’s dedication to proactive scam prevention. As AI capabilities evolve, we will continue finding new and innovative ways to harness this technology to protect our customers.
“Artificial intelligence is transforming how we protect customers,” Thomas Winkler, director of international trustworthy shopping experience said. “We can identify threats faster, act at scale, and build even greater trust in the shopping experience. As technology evolves, so will our ability to help customers stay safe around the world.”
We make it harder for bad actors to copy our communications with customers by employing industry-leading practices. If someone falls victim to a scam, they need to be able to quickly report it and understand how to move forward. Customers can report suspicious activity at Amazon.com/ReportAScam, available in 28 languages on mobile phones or desktop computers.
Learn more about the latest scam trends and how Amazon helps customers spot fraudulent content here.