AI in Marketing

Malicious Google Chrome extensions have compromised the data of nearly 900,000 users by mimicking legitimate AI-powered tools. Security researchers at Ox Security discovered that these extensions, posing as products from a company called AItopia, were designed to exfiltrate user data from popular large language models (LLMs) like ChatGPT and DeepSeek. The extensions, "ChatGPT for Chrome with GPT-5, Claude Sonnet & DeepSeek AI" and "AI Sidebar with Deepseek, ChatGPT, Claude and more," were able to collect complete conversation content and browser activity, sending it to a command-and-control server. Despite being featured on the Chrome store, these extensions have since been removed. The data stolen includes proprietary source code, business strategies, confidential research, and more, which can be used for corporate espionage, identity theft, and targeted phishing. The practice, known as "prompt poaching," raises significant privacy concerns as organizations increasingly rely on LLMs for sensitive tasks. Ox Security advises users to uninstall these extensions and be cautious of installing extensions from unknown sources. The monetization of this stolen data is facilitated by automated tools, making it easier to extract valuable information, including financial details and business credentials, which can be sold or used for targeted marketing and espionage.

In other news a marketing innovation

Omnicom has launched the next generation of its marketing intelligence platform, Omni, which integrates data, media, creativity, and AI to enhance marketing and sales strategies. This platform, built on the combined strengths of Omnicom and the recently acquired Interpublic, aims to provide a unified foundation for strategy, execution, and performance across the marketing ecosystem. Omni features a comprehensive identity and data infrastructure, including Acxiom RealID™ with 2.6 billion verified IDs and trillions of signals, and the largest media and commerce buying power globally. It also offers an end-to-end creative and content production infrastructure designed to scale personalization and performance. The platform supports decision-making with AI, providing insights and recommendations while maintaining human oversight. Omni's open architecture allows integration with existing marketing technology investments, ensuring flexibility and data ownership. The platform is designed to reduce complexity, bringing strategy, creativity, and execution closer together, enabling teams to make informed decisions and deliver measurable growth. Omnicom's CEO, John Wren, emphasizes that Omni's capabilities are enhanced when operating on a unified platform, breaking down silos and uniting teams to achieve precision and speed in marketing efforts. This development aims to connect every marketing element, from audience insight to sales activation, empowering brands to achieve significant growth.

From John Wren to CES territory

Retail media networks are increasingly central to advertising strategies, using first-party data and AI to enhance personalization and performance measurement. These networks are evolving from peripheral tools to core components of brand competition, integrating data across platforms to improve consumer experiences and marketing outcomes. At a CES panel, executives from Target, Meta, and Oura discussed how retail media is shifting focus from ad inventory to data-driven decision-making. AI supports productivity and strategic focus, aiding marketers in achieving better targeting and returns. The networks also extend beyond traditional retail, as seen with Oura's growth in wearable technology. The emphasis is on proving performance, with data enabling marketers to refine campaigns and meet rising consumer expectations.

Leaving CES for Paul Irving

TechCrunch reports that many well-funded startups struggle despite having strong products because they focus too much on development and not enough on distribution. Paul Irving, partner and COO at GTMfund, argues that in the AI-driven startup era, distribution is crucial. The traditional go-to-market strategies are outdated, and startups must innovate in distribution to succeed. Irving emphasizes that the speed of innovation has increased, requiring startups to differentiate their distribution methods rather than their products. He advises startups to adopt creative, data-driven approaches to reach their customers, suggesting that founders should not attempt to tackle all distribution channels simultaneously. Instead, they should focus on building a team of trusted advisers and approach hiring thoughtfully. GTMfund looks for startups that creatively reach their unique customers rather than those spending heavily on ads or large sales teams. Irving provides an example of a startup that effectively engaged with relevant Facebook groups to reach potential customers. He stresses the importance of a robust network, noting that GTMfund helps its portfolio companies by creating valuable connections. The venture-backed ecosystem is supportive, with many willing to help if approached with curiosity and a willingness to share knowledge. Irving concludes that founders should not work alone and must leverage their networks for success.

From Paul Irving toward Adam Kiefaber

Thought leadership in public relations often faces challenges in achieving widespread influence, according to Adam Kiefaber, senior director of content at Worldpay. Speaking at the PR Daily Conference, Kiefaber emphasized the importance of focusing on audience expectations rather than merely seeking media coverage. He noted that PR teams frequently prioritize what might gain coverage over what is genuinely useful to their audience. Kiefaber argued that understanding audience expectations should guide research and publication efforts. Worldpay, which processes payments in 174 countries, exemplifies this approach by concentrating on expertise in payment behaviors rather than commenting on every trending topic. This focus led to the creation of the Global Payments Report, a comprehensive analysis of customer behavior, now in its tenth year. Kiefaber suggested that this principle is applicable across various industries. For instance, HR firms can publish data on talent, while reputation companies can focus on trust data. The key is to align content with what the audience expects from the organization. This strategy not only enhances thought leadership but also builds a global influence. To access the full story and other archived content, readers are encouraged to become Ragan Insider members. The PR Daily News Feed offers the latest articles directly to subscribers' inboxes, ensuring they stay informed on industry developments.

From Adam Kiefaber to the Claude Code

AI's impact on productivity is causing mixed feelings among workers, particularly in software development, where tools like Anthropic's Claude Code and OpenAI's Codex are revolutionizing the field. These AI tools allow developers to complete tasks in minutes that once took days, leading to a significant shift in how software is created. While some developers are excited about these advancements, others feel a sense of loss as their hard-earned skills become less essential. The process of coding has shifted from writing code to managing AI tools, which some find disorienting. Gergely Orosz, a software engineer, expressed his struggle with AI's growing role in coding, noting that it performs tasks faster and often better than humans. Similarly, Andrew Duca, founder of Awaken Tax, shared his mixed emotions about AI's capabilities, acknowledging the efficiency it brings but lamenting the devaluation of his expertise. This phenomenon is not limited to software development; white-collar workers across various industries are likely to face similar challenges as AI continues to reshape professional landscapes. While AI can alleviate repetitive tasks, it also raises questions about the value of human effort and creativity. Some may find solace in managing AI effectively, while others might seek fulfillment in activities untouched by AI, like Erik Meijer, who returned to music to experience creative flow.

Claude Code gives way to Tsinghua University

Researchers from Tsinghua University and Peking University have developed DrugCLIP, an AI-driven screening program designed to accelerate drug discovery by efficiently scanning the druggable genome for potential ligands. The program addresses the challenge of computing power required for large-scale virtual screens, which traditionally take years to complete on a single server. DrugCLIP combines contrastive learning and dense retrieval, enabling cross-screening campaigns involving over 10 trillion protein–molecule pairs. This capability allowed the researchers to conduct a virtual screen of 10,000 human proteins and 500 million compounds in just one day, resulting in 2 million small-molecule hits. The program has been used to identify ligands for the serotonin 2A receptor, norepinephrine transmitter, and thyroid hormone receptor interactor 12, with the first two targets chosen for their clinical relevance in neuropsychiatric processes. DrugCLIP is freely available and is particularly suited for rapidly narrowing down large chemical spaces and prioritizing candidates. However, further work is needed to optimize the program, focusing on detailed pose generation, interaction constraints, and structure-level refinement. The development of DrugCLIP represents a significant advancement in drug discovery, making previously infeasible screening regimes practically achievable within a short timeframe.