Loading story…
A new study has demonstrated that even after regulatory restrictions on explicit ad targeting, algorithmic profiling remains prevalent on social media platforms such as Facebook. Researchers used a pipeline leveraging large language models (LLMs), specifically Gemini 2.0 Flash, to analyze over 435,000 Facebook ad impressions and found that these models could infer private user attributes—including age, gender, employment, education, and political preference—from ad exposure alone, with accuracy surpassing random guessing and demographic baselines. The study highlights that, despite Meta's removal of direct targeting options for sensitive categories, its ad ranking system still introduces demographic patterns that can be exploited by adversaries with access to visible ads.
Simultaneously, privacy advocates have raised concerns about Meta's new policy that personalizes ads and content based on users' interactions with its generative AI features across Facebook, Instagram, WhatsApp, and Messenger. The policy, which does not allow users to opt out, has prompted warnings that it could erode privacy protections, especially regarding sensitive topics discussed in AI chats. Experts question whether Meta's assurances about not using certain sensitive topics for ad targeting are sufficient, given the potential for indirect profiling and the use of proxy signals. The convergence of these developments underscores growing risks to user privacy from AI-driven ad targeting and profiling on major social media platforms.
See the reporting duties and controls this puts on the clock.
3 events from the most recent confirmed update back to the earliest known activity.
Researchers published a study showing that large language models could infer attributes such as age, gender, employment, education, and political preference from users' Facebook ad impressions, using analysis of more than 435,000 ads. The work found the technique could be scaled through browser extensions and may apply broadly to ad-driven platforms beyond Facebook.
Privacy advocates and experts publicly warned that Meta's new policy could weaken privacy protections, increase manipulation and scam-ad risks, and expose sensitive user information despite Meta's assurances about excluding certain topics from ad targeting. Critics also pointed to Meta's history of privacy violations and questioned the company's ability to reliably filter and audit such data use.
Meta implemented a policy allowing data from users' interactions with its generative AI features to be used to personalize content and advertising across Facebook, Instagram, WhatsApp, and Messenger. The policy drew concern because users reportedly could not opt out and may share sensitive information with the chatbots.
Vulnerabilities, threat actors, malware, products, organizations, and breaches Mallory has linked to this story.
See what this changes for your reporting obligations and which controls it puts on the clock.
2 references tracked. Mallory keeps watching after this page renders.
Map indicators from this story to your assets and identify affected systems in minutes.
Every observed campaign, victim, and pivot linked to actors named in this story.
Malware, exploits, and IOCs connected to the activity described here.
YARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Get matching new stories delivered to your team as they break — not the next morning.
Ask questions about this story and take action on the answers.
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
Meta changed its AI-generated suggested prompts after Facebook and Instagram user Kalie Robins reported that Meta AI asked invasive questions about her young daughters and assembled details from years of family posts. Prompts including “Who is the child passenger?” and questions about the children’s ages and the family’s residence reportedly led the chatbot to surface names, birth dates, photos, videos, and inferred location information, including an image Robins believed she had deleted years earlier. Meta said the personal-topic suggestions “missed the mark” and that it fixed the underlying prompt issue. The company maintained that Meta AI returns only material already accessible to the person making the query, but the incident renews scrutiny of the platform’s AI privacy controls following reports of publicly shared AI chats, an alleged private-chat ID-guessing exposure, and plans to use AI interactions for advertising targeting.
Sep 15, 2026
Meta is facing renewed scrutiny after the Tech Transparency Project reported that more than **7,500 ads** for AI-powered “nudify” services appeared on **Facebook** and **Instagram**, despite platform rules banning sexually exploitative content. The ads promoted tools that digitally remove clothing from images of women, raising concerns about AI-enabled sexual abuse and failures in ad moderation across Meta’s platforms. The report said the ad surge was tied to **GatherOne**, a Chinese advertising partner that may spend more than **$70 million annually** on Meta ads. Researchers also said one promoted app had previously been linked to **child sexual abuse material**. Meta said it takes aggressive action against nudify apps and can remove ads, impose financial penalties, or suspend advertiser accounts, but the findings are likely to intensify regulatory and public pressure over how the company handles abusive AI content and high-spending advertisers.
Aug 5, 2026
Two separate developments highlighted growing security and privacy concerns around consumer AI chatbots as the industry shifts toward ad-driven monetization. **Moxie Marlinspike** (Signal founder) launched *Confer*, positioned as an “end-to-end encryption for AI chats” tool intended to keep user conversations private—claiming the service cannot read chats, store them, or use them for model training because only the user can access the plaintext. The reporting frames *Confer* as a response to increasing workplace and consumer leakage of sensitive data into AI assistants and broader concerns that AI platforms will monetize user data. In parallel, commentary warned that major AI providers are increasingly adopting the social-media model of **behavioral data collection and advertising**, raising the risk of user manipulation and erosion of trust. The piece cites OpenAI’s expansion into search and browsing features and notes OpenAI’s announcement that it will begin testing ads in the free version of **ChatGPT**, alongside similar ad experiments by other AI products (e.g., Perplexity, Microsoft Copilot, Google’s AI search experiences, and Amazon’s chatbot). The core security implication is that ad incentives can drive more tracking and persuasive optimization in AI responses, increasing privacy exposure and the potential for covert influence over user decisions.
Jun 29, 2026