Researchers find Israeli spy company targeted Starlink users.

7-minute read | 1,600 words

Alleged Israeli spy company targeted Starlink users.

Telemetry data shows that Discover-Tech was the number one bidder on Starlink traffic in over one hundred countries and looked to geolocate or surveil users.

US political candidates tailoring messages for AI models.

Political candidates have been increasingly adjusting their messaging to appeal to AI chatbots that voters are engaging with.

Researchers find Israeli spy company targeted Starlink users.

THE NEWS

On Tuesday, researchers at TrendAI alleged that Discover-Tech, an Israeli company, was one of the largest sources of digital advertisements shown to Starlink’s users. Furthermore, researchers found that these advertisements were likely being used to geolocate or surveil users. 

Through their research, TrendAI found that Discover-Tech had become one of the leading bidders for Starlink user ads, taking the number one bidder spot for Starlink traffic in over 100 countries. 

When publishing their report on October 15th, TrendAI found Discover-Tech, labeled as Shadow-Earth-79, was able to abuse the ad bidding process by leaking data about Starlink users. Over time, the bidder can collect, collate, and merge leaked data with other information not only deanonymizing users but also following them and potentially targeting them with surveillance software.

TrendAI commented that Shadow-Earth-79 was:

“Targeting and tracking infrastructure rather than a commercial advertising operation.”

Additionally, TrendAI cited an “obvious skew” towards buying ads targeting Starlink users obtaining “an implausibly dominant share” of bids in countries including Hungary, Israel, Egypt, Azerbaijan, and Venezuela.

In response to the previous investigations, Google and Comcast have removed Discover-Tech from their ad platforms.

THE KNOWLEDGE

While the use of advertising infrastructure for surveillance operations is not new, the alleged targeting of Starlink users highlights how commercial advertising can be repurposed for intelligence gathering. Further, this is not the first time Israeli companies have been linked to spyware operations.

Several years ago, the NSO Group came under global scrutiny due to their use of the Pegasus spyware. With this software, the company targeted journalists, activists, and other individuals covertly compromising mobile devices to gain access to sensitive information, messages, location data, and contact information. 

This latest incident highlights a different method of identifying and compromising targets. In a separate investigation, Citizen Lab linked Discover-Tech’s advertising operations to efforts to deliver spyware to users. Specifically, Discover-Tech was able to deliver Israel’s spyware company Candiru software on its behalf to users reaching over 120,000 websites and over 75,000 applications. In some cases, advertisements could collect information and track users simply by loading without needing user engagement.

New York advertising security firm, Confiant, noted that these spyware-laced ads likely impacted hundreds of millions of people. 

Together, the various tactics employed by Citizen Lab illustrate how commercial advertising can serve as more than a vehicle that delivers ad content. By combining collected user data with targeted advertising spyware, systems can provide surveillance operations with ways to identify, track, and potentially compromise individuals through routinely used services.

THE IMPACT

These findings raise significant questions about the security of digital advertising and the extent that third-parties can expose users to surveillance. While organizations invest in securing networks and devices, advertising platforms create additional exposure that are often outside of someone’s ability to control.

For Starlink users, the consequences extend beyond privacy. Starlink provides valuable connectivity across conflict zones and where internet access is limited. However, if advertising data can be used to locate or profile individuals, that could expose their activities and movements, creating numerous security risks.

Notably, this risk extends beyond any single advertising company or satellite provider. As surveillance operations identify new ways to exploit commercial infrastructure, organizations will need to account for the associated security risks. Protecting sensitive users will require not only securing devices and networks, but also understanding and improving oversight over the systems that passively collect, process, and distribute information.

AI chatbots are shaping political messaging.

THE NEWS

On Monday, reports emerged detailing how US political candidates are changing their campaign messaging strategies to target AI chatbots. With this pivot, candidates are looking to engage with the chatbots that voters have been increasingly using to answer questions on various election topics. 

For example, Mike Minogue, a Republican gubernatorial candidate for Massachusetts, was given a strategy that looked to convince ChatGPT, Claude, Gemini, and other models that he was the preferred choice on key issues. The strategy looked to create hundreds of AI-friendly policy pages and citations across major platforms scraped by models and then alter content accordingly until answers reflected the campaign’s platform.

Additionally, the strategy was looking to shape the information environment on how models describe, compare, and recommend candidates. More specifically, the strategy looked to buy placements on high-authority sites commonly cited by models and produce YouTube and Reddit content for models to pick up. The strategy wrote:

“When the campaign is ready, we run paid placements on high-domain-authority sites the LLMs already cite – plus Reddit, which is disproportionately weighted in retrieval.”

Joe Bonilla, managing partner of Relentless Awareness, commented on the trend, noting:

“Campaigns using AI-influence training are now a reality that we face and it will be ubiquitous by the 2028 cycle. The public increasingly cannot and will not be able to discern what is created by systems like Claude and ChatGPT versus human-originated content.”

THE KNOWLEDGE

As AI has become a more universal tool, it is not shocking that political campaigns have increasingly turned to it for a variety of purposes. Outside of using them to inform voter opinions when engaging with chatbots, campaigns have also begun to use them when constructing advertisements. 

Michael Franz, the co-director of the Wesleyan Media Project, has tracked the emergence of these ads. Franz noted how deepfake ads are “getting way more common” documenting over 160 ads being either produced or enhanced by AI across the nation. 

Rather than deceptive, these AI ads are being used satirically, depicting political opponents and creating humorous and emotionally persuasive content at scale for affordable costs.

Todd Stiefel, a backer for Democratic candidates in North Carolina, commented that their use of AI ads allowed them to point out additional overlooked information. Stiefel commented:

“We decided that AI could be a way that we could help people see the words that came out of his mouth, although technically they came out of his typing.”

This increasingly multi-faceted use of AI in campaigns creates new opportunities for political messaging, allowing candidates to produce content more affordably while still influencing how voters understand and engage with issues. Whether through AI- generated content or chatbot response engagement, these strategies are looking to place messaging in the environments that voters are increasingly likely to spend their time.

However, these approaches raise questions about how voters are able to distinguish real information from persuasive content. By warping chatbot responses to sway opinions or creating realistic AI ads, campaigns turn AI into a tool that can be used for information gathering and analysis into a propaganda device that does not consider if the information being presented is accurate.

THE IMPACT

The growing use of AI in political campaigns is already changing how candidates compete for voter attention and how they shape public opinion. As voters continue to turn to chatbots for information, the responses these systems generate will likely become another point of influence for political campaigns. Efforts to shape response through source manipulation may prove to be a major campaign investment point, similar to the introduction and widespread use of social media over a decade ago.

However, the implications of AI use extend beyond chatbots. AI-generated advertisements can allow campaigns to create persuasive content quickly and at scale. While many of these satirical ads are not deceptive, they blur the lines between fabricated scenes, altered statements, and authentic recordings. 

As campaigns continue experimenting with these tactics, the challenge will be ensuring voters can evaluate political messaging without assuming that AI-generated content is accurate or impartial. AI may expand the ways campaigns communicate, but it also creates a greater need for transparency, source verification, and voter discernment.

This Week’s Caveat Podcast: The AI election.

Dave Bittner and Ben Yelin break down how political campaigns are increasingly using AI to engage with voters and another story of a woman who faces years in prison after being mistaken for a scammer who stole her identity. Afterwards, Ben sits down with Gary Barlet, Illumio’s public sector CTO to discuss election security.

Listen to the episode now. 

OTHER NOTEWORTHY STORIES

Anthropic model accessed voter information.

What: Claude was able to bypass a paywall to obtain Georgia voter count data without authorization.

Why: On Sunday, Anthropic released a statement confirming that its Claude AI model was able to bypass a paywall to access Georgia’s voter information without human authorization. In the statement, Anthropic wrote:

“The model was being used internally for a statistics project to pull public data. Without being asked to, the model requested a token to query a dashboard that contained information normally available for a fee. We have since put in place mitigations to block these kinds of actions.”

Anthropic has notified the Georgia secretary of state’s office and the White House of the incident. 

OCTOBER 25, 2026 | Source: Axios

Texas sues OpenAI.

What: Texas attorney general filed a lawsuit against OpenAI alleging product safety concerns.

Why: On Friday, Texas Attorney General Ken Paxton filed a lawsuit alleging that OpenAI had misled consumers about the safety of its AI chatbot and that the company failed to protect users from harm.

When filing the lawsuit, Paxton stated:

“With the rapid advancement of AI technology, OpenAI has the responsibility to protect users and warn them of the many dangers of its platform.”

The lawsuit alleges that OpenAI violated the state’s consumer protection and children’s online safety laws.

OCTOBER 23, 2026 | Source: Politico