Where is AI hiding inside WhatsApp



Image by Rajashree Rajadhyax


WhatsApp is one app that we use everyday. I’m sure with many that is the default way of staying connected with the world. Its easy to use, always available and the best thing is that its associated with your mobile number. I think the success behind WhatsApp over any other messaging tool is its association with the mobile number. While the interface is clean and simple and there is no bragging about use of AI, yet AI is working in the background to make our WhatsApp experience smooth. Let’s do a deep dive.


Smart and optimised message delivery

When you send a message on WhatsApp, it feels simple and instant. But real-world networks are messy. Signals drop, phones switch between Wi-Fi and mobile data, battery savers kick in, and the person on the other side may be offline. Yet messages usually do reach. This is because WhatsApp doesn’t just send a message once and forget about it. That would be a purely programmed or deterministic behaviour. Instead, when the network is weak, you see the small clock icon indicating that the message is held back. WhatsApp waits, observes how the network behaves, and retries at the right moment. When conditions improve, delayed messages often go through together, suddenly turning into double ticks.

Over time, WhatsApp learns how messages behave across different networks and regions. It learns when retries usually succeed, when waiting works better, and how often to try again without wasting battery or data. At the same time, WhatsApp operates a massive global infrastructure, and messages do not all take the same path. Based on learned patterns, the system chooses better delivery routes, avoids overloaded servers, and balances traffic across regions so messages can move more smoothly. All of this is done using network signals, timing, and delivery outcomes, not by reading the message content. End-to-end encryption remains intact, your messages are not seen.

WhatsApp does not publicly publish technical papers calling this “AI,” but this kind of adaptive, experience-based decision-making is well understood in large-scale systems as something that cannot be achieved with fixed rules alone. It is therefore reasonable to infer that learning-based, AI-driven systems are quietly working in the background to make message delivery feel reliable, even when the network is not.


Spam, Scam, and Abuse Detection

If WhatsApp can’t read your messages because they are end-to-end encrypted, then how does it detect spam, scams, or abuse at all?
WhatsApp cannot see message content such as the text, images, or voice notes are encrypted and private. But it can observe behaviour around messages: how many people an account messages in a short time, how often messages are forwarded, how quickly a new account starts bulk messaging, and how recipients react by blocking or reporting. These patterns exist outside the message itself, and they are enough to spot behaviour that looks very different from normal human conversation.

This kind of detection cannot work well with fixed, hard-coded rules alone, because spam and scam tactics constantly change. Instead, WhatsApp uses automated systems that learn from past behaviour such as what spam looked like before, how scams spread, and which patterns repeatedly lead to reports and blocks. While WhatsApp does not publicly publish detailed technical papers explicitly describing these systems as “AI”, it does confirm the use of automated spam detection, and independent reporting and past disclosures strongly indicate learning-based, pattern-driven systems at work.

Image, Video, and Media Compression

Every time you send a photo or video on WhatsApp, something interesting happens before it ever leaves your phone. The app doesn’t just shrink the file using a fixed formula. It tries to understand the media first, for example, what matters in it and what doesn’t etc and then compress it in a way that keeps it usable while reducing size.

For images, WhatsApp needs to decide where it can afford to lose detail. Faces, text, and sharp edges matter more to humans than flat backgrounds or noisy areas. Modern compression systems increasingly rely on models trained on human perception to preserve what people notice most, while compressing less important regions more aggressively. The result is familiar: images that look “good enough” on a phone screen, but are much smaller than the original.

Video, bandwidth, and real-time decisions

Video makes the problem harder. A video has motion, changing scenes, and varying levels of detail. WhatsApp adapts video quality based on network conditions, device capability, and expected playback quality. AI-style systems help predict how much a video can be compressed before people start noticing artifacts like blur or blockiness. They also help decide when to lower resolution, drop frames, or reduce bitrate so the video keeps playing smoothly instead of stalling.

At WhatsApp’s scale, these decisions cannot be fixed or one-size-fits-all. Networks behave very differently across regions, and phones range from basic low-end devices to powerful high-end ones. WhatsApp does not publish technical papers saying “this is our AI model for media compression,” but the way the system behaves tells its own story. The quality adjusts automatically, important visual details are preserved more than background noise, and the app makes smart trade-offs based on network conditions. This closely matches how modern image and video systems use machine-learning techniques today. And once again, the goal is not perfect quality. It is fast delivery, low data usage, and media that looks good enough on the device in your hand, quietly optimised so you never have to think about it.

Voice and Video Call Quality

When you make a voice or video call on WhatsApp, the app is constantly reacting to what is happening on the network in real time. If bandwidth drops, WhatsApp may lower audio bitrate, reduce video resolution, or prioritise voice over video so the conversation stays understandable. If packets are lost, it uses techniques like buffering, error correction, and noise handling to avoid sudden gaps or robotic sound. These decisions are not made once at the start of the call. They keep changing as network conditions change. After the call ends, WhatsApp sometimes asks you to rate the call quality. That feedback becomes a signal. Over time, patterns emerge between network conditions, technical adjustments, and how users actually experience the call. WhatsApp does not publicly share its internal AI models for call quality or account security. However, the system’s adaptive voice and video quality adjustments such as real-time bitrate and resolution changes are consistent with techniques used in machine-learning-guided optimisation of real-time communication systems.

Security Signals and Account Protection

Have you noticed that sometimes WhatsApp may ask you for additional verification information, or suddenly restrict certain actions even though you feel you have done nothing wrong? This usually happens when something about your account activity looks unusual. It could be a login from a new device, repeated verification attempts, a sudden change in usage patterns, or behaviour that does not match how you normally use WhatsApp. Instead of treating every login or action the same way, WhatsApp assesses risk in real time and decides when to add extra checks to protect your account.

This is where AI quietly plays a role. WhatsApp learns what “normal” behaviour looks like for accounts over time and flags situations that deviate from those patterns. Based on these signals, it may ask for extra verification, temporarily limit certain actions, or prompt security steps like two-step verification. WhatsApp does not publish technical details describing these systems as AI, but the behaviour clearly goes beyond fixed rules. Learning from patterns, adapting to new threats, and responding differently based on risk levels are all characteristics of AI-driven security systems designed to protect users without constantly interrupting them.

Final thoughts

WhatsApp’s use of AI is not about flashy features or obvious buttons. It is about making everyday things work smoothly, without demanding your attention. Messages reach you even when networks are unstable. Calls stay clear as conditions change. Spam and scams are pushed out before they spread. And all of this happens while keeping data usage low and privacy intact.

Most of the time, you don’t stop to think about why WhatsApp works so reliably. And that is exactly the point. The best AI systems are the ones you never notice. They don’t announce themselves. They quietly learn from patterns, adapt to real-world conditions, and step in only when needed.

In the next article, I’ll come back with another everyday app we all use, and discuss where AI is quietly working behind the scenes there as well.

References

  1. https://www.braze.com/docs/user_guide/message_building_by_channel/whatsapp/whatsapp_campaign/optimized_delivery/#setting-up-optimized-delivery

  2. Spam detection: https://www.whatsapp.com/security

  3. https://www.leapxpert.com/how-whatsapps-new-security-features-make-blocking-spam-much-easier/

  4. https://internetlab.org.br/wp-content/uploads/2019/10/WA_StoppingAbuse_Whitepaper_020618-Final-1.pdf?utm_source=chatgpt.com

  5. https://www.bitdefender.com/en-us/blog/hotforsecurity/meta-adds-anti-scam-features-messenger-and-whatsapp

Comments

Post a Comment

Popular posts from this blog

5 Skills to Work Better with AI

Image
  AI is everywhere, and it’s here to stay. From virtual assistants to powerful decision-making tools, AI is shaping the way we work and interact with technology. But simply using AI isn’t enough—you need to understand how to make the most of it. Many people, even in everyday interactions with AI-powered tools like ChatGPT, Gemini, or DeepSeek, struggle to get the results they expect. This isn’t because AI is flawed, but because working effectively with it requires a certain set of skills. Knowing how to phrase your queries, provide the right context, and interpret AI’s responses can make all the difference. Whether you use AI for work, creativity, or problem-solving, keeping these five key skills in mind will help you unlock its true potential. 1. Know Where to Use AI AI isn’t a one-size-fits-all solution. While it excels at processing vast amounts of data and analyzing lengthy documents, it’s not always necessary—or even useful—in every situation. It’s not a magic wand, and unders...
Read more

AI learns language with the help of a child

Image
  Today I want to talk to you about an experiment conducted by some AI researchers to see if AI can learn language the way human babies do. With this in mind, they put a head mounted camera on an infant and tried to understand how human babies comprehend language with the help of the visual and auditory cues it gets from its environment and can AI learn on the same lines. We’ll discuss that experiment in a bit. If you come to think of it, learning a language is such a complex task and yet human babies do it pretty easily! Contrary to the belief that parents or caretakers teach a child to learn a language, children acquire language through interaction — not only with their parents and other adults, but also with other children. All normal children who grow up in normal households, surrounded by conversation, will acquire the language that is being used around them. And it is just as easy for a child to acquire two or more languages at the same time, as long as they are regularly int...
Read more

Part 1: Prompt engineering: The Art of Delegating to Generative AI

Image
Ever wondered why some people get spot-on, helpful responses from AI, while others end up with vague or confusing answers? It’s not just luck—it’s the power of good prompting. The way you ask the AI to do something matters—a lot. This is part one of a three-part series where we’ll explore what prompt engineering is and the different techniques you can use to guide AI effectively. Part two will cover the constituents of a prompt. In part thre, we’ll dive into how to create great prompts—with practical tips and examples to help you get the most out of AI. What is prompt engineering? So, what exactly is prompt engineering? It’s the art of telling an AI model what you want it to do. But why is it called an art—and yes, also a science? Because the way you frame your request can completely change the quality, accuracy, and usefulness of the response. To understand why prompting is such a big deal, you first need to know how these AI models actually work. The popular chat companions we’ve all...
Read more