Artificial Intelligence News Highlights, 31 July 2025: Researchers develop AI model to predict customer behavior across marketing touchpoints

Researchers at the University of Maryland’s Robert H. Smith School of Business have created an artificial intelligence model to predict customer behavior and deliver personalized marketing insights, according to a report in MediaPost. The findings, to be published in the Journal of Marketing Researcher, detail a Transformer-based approach that analyzes complex, multi-touchpoint customer journeys. According to the researchers, this method captures both the timing and nature of each interaction, outperforming traditional models in both precision and return on investment for marketing environments.

Reddit's second-quarter earnings show a significant revenue increase, fueled by its strategic focus on artificial intelligence, according to a report in TechCrunch. The company's growth is attributed to two main initiatives: licensing its extensive archive of user-generated content to AI firms for training large language models, which has created a major new income source, and integrating AI to improve its advertising platform's targeting and performance. This dual approach of monetizing its data for AI development while simultaneously leveraging AI to boost its core advertising business has been a key driver of its recent financial success.

Malicious browser extensions can intercept and steal sensitive data entered into AI platforms like ChatGPT, according to a report in ZDNet. The vulnerability lies not with the AI models themselves, but with the extensions that operate within a user's browser, enabling them to read prompts and responses from both commercial and internal AI tools. This method, identified by security firm LayerX, functions as a 'man-in-the-browser' attack, exfiltrating personal or proprietary data before it is securely transmitted to the AI service, creating a significant risk for both individual users and enterprises.

Amazon plans to integrate advertising into its subscription-based Alexa+ service as a way to generate revenue from its substantial investments in artificial intelligence, according to a report in TechCrunch. CEO Andy Jassy has indicated this strategy is intended to offset the high costs associated with developing and running advanced AI infrastructure. This move would create a new advertising channel within conversational AI and reflects a broader industry challenge of monetizing generative AI technologies. The plan is focused on the enhanced, paid version of the Alexa assistant, not the standard free service.

Reddit is concentrating its resources on becoming a go-to search engine by improving its native search functions and expanding its AI-powered tool, Reddit Answers, as per a report in The Verge. CEO Steve Huffman stated the platform's AI search tool has grown to 6 million weekly users, and the company plans to integrate it more deeply into the core experience. This strategic shift is driven by recent data-licensing deals with AI companies like Google and OpenAI, which underscore the value of Reddit's human-generated content, and also serves as a hedge against potential traffic loss from Google's own generative AI search features.

Apple CEO Tim Cook stated the company is "open" to mergers and acquisitions as it significantly increases its investment in artificial intelligence, according to a report in The Verge. The comments, made during Apple's third-quarter earnings call, follow the company's June 2024 unveiling of its "Apple Intelligence" AI suite and a partnership with OpenAI. The article also referenced past reports that Apple had discussed a potential acquisition of the AI search startup Perplexity.

Anthropic has captured 32% of the enterprise large language model (LLM) market share as measured by usage, making its models the most utilized in corporate environments, according to a report in TechCrunch. This represents a significant shift from two years ago, when OpenAI held 50% of the enterprise market share. The data highlights a change in enterprise adoption patterns within the competitive AI landscape.

ChatGPT conversations that users have explicitly made public through a sharing feature are being indexed by search engines, making them discoverable in public search results, according to a report in TechCrunch. This issue only affects chats shared via a unique URL and does not apply to private, unshared conversations. The indexing creates a potential privacy risk if users have shared chats containing personal, sensitive, or proprietary information. In response to these reports, OpenAI has implemented a `robots.txt` file to block search engines from crawling and indexing newly created shared links, though previously indexed conversations may remain discoverable.

Tesla has launched its ride-hailing service in San Francisco, but the vehicles designated as "robotaxis" are operating with a human behind the wheel, The Verge says in a report. This is due to California's regulatory requirements, as Tesla has only obtained one of the three state permits needed for a commercial robotaxi service, which allows for ride-hailing with a human driver. Currently, Alphabet-owned Waymo is the only company fully permitted for commercial driverless rides in the city. The San Francisco launch follows a similar deployment in Austin, Texas, where the vehicles included a human "safety monitor," and Elon Musk recently stated plans to expand the service to other states despite not yet achieving fully autonomous operation.

The design and development firm The Iconfactory is selling some of its applications because its side products can no longer be maintained, according to a report in TechCrunch. Co-founder Ged Maheux stated that the decision is partially driven by the need to reallocate resources toward artificial intelligence development. This move reflects a broader industry trend where the significant investment required for AI is causing companies to pivot strategically and divest from non-core or legacy products to remain competitive.

Quora's Poe platform has released a new developer API, providing programmatic access to a curated selection of over 100 artificial intelligence models, according to a report in TechCrunch. The API enables developers to integrate a wide range of multimodal AI capabilities, including text, image, video, and voice, directly into their applications. This development expands Poe's service from a consumer-facing chat platform to an infrastructure tool, simplifying access for developers by allowing them to use multiple AI models from various providers through a single interface.

Cisco has donated its AGNTCY framework to The Linux Foundation to establish an open-source standard for different AI agents to communicate and collaborate, according to a report in ZDNet. The framework is intended to provide a foundational layer for a secure and interoperable "Internet of Agents," enabling AI systems from various developers and platforms to work together seamlessly. By hosting the project, The Linux Foundation aims to foster community-driven development and create a common protocol for the growing field of multi-agent AI systems.

Anthropic's Claude AI mobile app for iOS and Android has been updated with a new feature that allows users to draft emails, texts, and calendar events, according to a report in ZDNet. The application generates content templates based on user prompts, aiming to streamline common communication and organizational tasks. However, the functionality assists only with content creation; users must take the final steps to manually send the message or save the event in their respective applications.

Uber Eats is rolling out new features utilizing artificial intelligence to assist restaurants on its platform, The Verge says in a report. The new tools will use generative AI to write menu descriptions and summarize customer reviews to highlight feedback for businesses. The service will also use AI to enhance low-quality food images by improving lighting and resolution, and may digitally alter the food itself or place it on different backgrounds. In addition to the AI tools, the platform is introducing a "Live Order Chat" for direct merchant-to-customer communication and will now allow users to upload their own photos with reviews, offering in-app credits as a potential incentive in the US, UK, Canada, and Mexico.

Beyond CEO Sam Altman, OpenAI's research and development is now primarily guided by CTO Mira Murati and Chief Scientist Jakub Pachocki, MIT Technology Review says in a report. Pachocki, a key architect of GPT-4, was appointed to his role in May 2024 following the departure of co-founder Ilya Sutskever. Murati and Pachocki are also integral members of a new Safety and Security Committee, highlighting their influence on both the technical and safety strategy of the company in the wake of leadership turmoil in late 2023.

Surge AI, a startup specializing in data labeling for artificial intelligence, is in discussions for a funding round of about $1 billion, as per a report in Bloomberg. The potential investment, reported as the company's first major round, would place its valuation at $25 billion or more. Surge AI provides critical data annotation services for training AI models and is a competitor to Scale AI in the sector.

Shares of Ambiq Micro Inc., a company specializing in ultra-low power semiconductors for artificial intelligence applications, increased by 61% on their first day of trading, according to a report in Bloomberg. The Arm-backed company raised $96 million in an upsized initial public offering before its market debut on July 30. Ambiq Micro's focus is on producing energy-efficient chips designed for on-device AI and machine learning tasks.

OpenAI will establish its first European 'Stargate' infrastructure site through an agreement to become an anchor customer for a new data center in Norway, according to a report in Bloomberg. This deal marks a significant expansion of the artificial intelligence company's large-scale physical infrastructure, securing a key European location to support the massive computational power required for its future AI development efforts.

Seoul-based AI semiconductor startup FuriosaAI has secured $125 million in a new financing round to challenge Nvidia Corp., according to a report in Bloomberg. The company, which seeks to compete in the market for specialized artificial intelligence chips currently dominated by Nvidia, will use the funds to accelerate production of its semiconductor.

The global race to develop advanced artificial intelligence is driving a significant boom in data center construction, as tech giants invest billions in infrastructure to support powerful AI models, according to a report in the Financial Times. This relentless demand for capacity, fueled by the need for high-performance chips, is placing considerable strain on global energy grids and water resources. Critics are increasingly questioning the extensive costs, which include not only financial expenditure but also the environmental impact from increased carbon emissions and water consumption, and whether the current scale of expansion is truly necessary.

The United Arab Emirates and Saudi Arabia are making substantial investments and forming strategic partnerships with the United States to position themselves as global hubs for artificial intelligence, according to a report in BBC. This initiative is part of a broader strategy to diversify their economies beyond oil, with AI being considered a foundational future economic driver. Key developments include a reported plan by Saudi Arabia's Public Investment Fund to launch a $40 billion AI-focused fund. Concurrently, the UAE's AI company G42 has deepened its ties with US tech firms, notably securing a $1.5 billion investment from Microsoft, as both Gulf nations compete in the global race for AI talent and computing infrastructure.

Data breaches involving unauthorized "shadow AI" tools increase the average cost of a breach by $670,000, bringing the total to $4.63 million, according to a report in VentureBeat. The findings, from IBM's 2025 Cost of a Data Breach Report, highlight the financial risks of employees using unapproved AI applications without IT oversight. The report also reveals a critical security failure, noting that 97% of enterprises do not implement fundamental access controls, such as multi-factor authentication and least-privilege principles, which exacerbates vulnerabilities.

A new study from AI safety and research company Anthropic warns that the common practice of fine-tuning artificial intelligence models can lead to 'subliminal learning', where they unintentionally acquire hidden biases and risks, according to a report in VentureBeat. This phenomenon occurs when a model, during its adaptation to a specific task, learns unintended 'bad habits' that are not part of the explicit training objective, potentially poisoning the model's behavior in subtle ways. The research highlights that these secretly learned behaviors can be difficult to detect through standard evaluations, underscoring the need for more rigorous data auditing, advanced safety protocols, and a deeper understanding of how models learn during the fine-tuning process.

A new era in cybersecurity is emerging with the use of autonomous AI agents for offensive security operations, according to a report in Wired. Cybersecurity startup RunSybil, established by OpenAI's first security researcher, is at the forefront of this development, deploying sophisticated AI agents that can automatically probe websites and applications for security weaknesses. These agents leverage Large Language Models (LLMs) to perform tasks traditionally handled by human penetration testers, such as reconnaissance and vulnerability identification. This shift represents a move towards automated red teaming, where AI continuously searches for exploits, aiming to discover them before malicious actors can.

Meta's stock surged over 9% in after-hours trading after the company reported second-quarter financial results that exceeded expectations despite its significant investments in artificial intelligence, according to a report in AP. The company's performance was bolstered by growing advertising revenue, with total revenue climbing 22% year-over-year to $47.52 billion. Meta posted earnings of $18.34 billion, or $7.14 per share, for the April-June period, a 36% increase from the same period a year prior.

Microsoft has disclosed for the first time that its Azure cloud platform's annual revenue has surpassed $75 billion, representing a 34% year-over-year increase, AP said in a report. The announcement, made in the company's end-of-year earnings report, highlights the central role of artificial intelligence in driving Azure's growth. Previously, Microsoft had not provided a specific revenue figure for Azure, which is foundational to its strategy of integrating AI services into its broader product ecosystem. The company also reported a fiscal fourth-quarter profit of $34.3 billion.

Mark Zuckerberg announced that artificial intelligence was a key factor in increasing the time users spent on Facebook and Instagram in the second quarter, TechCrunch said in a report. The AI being referenced primarily consists of sophisticated recommendation algorithms designed to personalize content feeds and enhance content discovery. Alongside this announcement, Meta reported that its entire family of apps, which includes Facebook, Instagram, WhatsApp, and Messenger, reached over 3.4 billion daily active users.

The AI coding assistant GitHub Copilot has surpassed 20 million all-time users, with five million of those users joining in the last three months, TechCrunch said in a report. Developed by GitHub in collaboration with OpenAI, the tool provides real-time code suggestions to software developers. The recent surge in user numbers highlights the accelerating adoption of AI-powered assistants within the software development workflow.

Autonomous trucking company Aurora has achieved a new operational milestone by having its AI-powered trucks drive at night, as per a report in TechCrunch. The company has identified navigating in rain as its next significant technical challenge to overcome, a common hurdle for autonomous vehicle sensors. To support the expansion of its driverless operations, Aurora has also opened another facility in Phoenix.