OpenAI claims ChatGPT reaches 20% of Google's search volume

Google made a major announcement that one of its Android partners intentionally installed malware on 1.3 million devices. Subsequently, a botnet was created through which users' devices and their internet connections were secretly used.

According to the report, this illicit activity was carried out by the manufacturer VideoStrong Technology, based in Shenzhen, China. Google specified that this malware was primarily found in Android TV set-top boxes. With this software, the company obtained permission from its 'criminal accomplices' to use users' internet connections without their knowledge.

In its blog, Google stated that it is fully confident that VideoStrong Technology developed and installed this backdoor and botnet for financial gain, and subsequently controlled them.

The company emphasized that this was not the result of an error by any single employee or oversight during development. Consequently, Google banned VideoStrong Technology, which can no longer create certified Android devices or provide software for them.

VideoStrong Technology is a company involved in Original Equipment Manufacturing (OEM) and Original Design Manufacturing (ODM). In addition to preparing products for its clients, it also releases its own products. The company was founded in 2011, and according to its website, its products have reached over 100 million households in more than 60 countries.

Google explained that a botnet is a network of infected devices. Attackers can take control of a device after the malware is installed. Infected devices are used for advertising fraud, data theft, and malware installation.

Most of the infected devices were Android TV set-top boxes, but the malware was also found in telemedicine devices, karaoke machines, and digital signage systems. The company also took legal action to disable the domains and servers controlling the botnet.

This case is not the first; similar incidents have occurred before. Last year, German authorities blocked a botnet called BadBox, which affected millions of Android devices. Furthermore, another botnet named Vo1d infected about 1.6 million Android televisions.

Although AI Mode is encouraging users to perform more searches on Google, it still results in lower revenue per query compared to conventional search. Alphabet confirmed that queries made through the artificial intelligence-based tool have a lower monetization rate than traditional searches.

This information was disclosed during the presentation of third-quarter results. Revenue from searches and other ads grew by 12% year-over-year, reaching US$ 63.3 billion (approximately R$ 338.7 billion in direct conversion). However, this growth rate was slower compared to the 17% recorded in the second quarter.

Philipp Schindler, Chief Business Officer, clarified that part of this disparity lies in the format of the queries themselves. Many of these interactions occur in preliminary stages of the consumer decision-making process, implying that not all of them have a strictly commercial objective.

According to Google data, a search conducted in AI Mode contains, on average, three times more words than a standard search. This feature is capable of processing complex questions and allows the user to maintain a continuous dialogue with new requests.

AI Mode is available in Portuguese and provides an experience similar to chatbots like Gemini and ChatGPT. Unlike a simple list of links, this tool breaks down questions into multiple searches and constructs an answer using information collected from the internet.

The strategy to improve revenue involves inserting more advertising into the tool. Google began testing advertising in AI Mode in May 2025, presenting sponsored links that can be displayed within or below the answers generated by artificial intelligence. The company plans to expand these ads to Australia, Canada, India, and seven other markets this week. Furthermore, Brazil, Mexico, and the United Kingdom are expected to be included in this expansion by the fourth quarter of 2026. In the United States, Google has also begun testing ads displayed during questions within the same conversation.

Search remains one of Alphabet's main revenue sources. Although the segment showed quarterly advancement, the 12% growth fell short of the 17% in the second quarter and the 19% in the first quarter. The lower monetization of AI Mode is an additional point of concern, given that Google is investing more resources to maintain its artificial intelligence strategy.

Alphabet's infrastructure spending reached US$ 51.4 billion in the quarter, equivalent to over R$ 275 billion in direct conversion. This amount intensifies the need for new AI products to help sustain the company's advertising business.

Nevertheless, Google shows no intention of slowing down. The company assures that AI tools are already increasing search volume and plans to further increase its technology investments in 2027. The main current challenge is making these new searches as profitable as previous modalities.

OpenAI introduced a new privacy feature in ChatGPT this Tuesday, October 20th: locked conversations. Users of OpenAI's artificial intelligence now have the option to mark sensitive interactions so that they are hidden from the service history.

The content of these protected conversations is secured by a PIN and can also be accessed via biometrics on a smartphone. However, OpenAI clarifies that locking interactions does not constitute complete security protection; there is no end-to-end encryption, and the way the company can access and use the information remains unchanged. Furthermore, ChatGPT continues to use these conversations as a basis for generating future responses.

Although this feature is not inherently a security measure, it serves to prevent curious people from viewing the content of questions asked to the AI. It also offers privacy to users with Business accounts, whose workspace administrators cannot reset the PIN.

OpenAI warns that if the user forgets the configured PIN, resetting this code will result in the permanent deletion of the blocked interactions, an action that is irreversible, regardless of whether the message was exchanged before or after closing the conversation.

The established PIN is unique to the entire account, and the chat itself requests the creation of this key the first time a conversation is locked. After setting the four digits, this password applies to all interactions protected by the feature. In situations of multiple consecutive errors when entering the PIN, the conversations are temporarily blocked to prevent third-party access.

For users who lose access to the correct PIN and do not have facial or digital recognition features, the only alternative to lock a new conversation is to reset the key, which implies the loss of previously protected content. However, changing the key itself does not lead to file loss, provided the previous PIN is remembered, allowing only substitution with an easier-to-memorize sequence.

The locked conversations feature is available for most ChatGPT modes and subscriptions and will be gradually implemented for all users. It is important to note that there are exceptions, as conversations are not prevented from being opened in contexts related to legal or compliance issues.