Data For Sale
During a Senate hearing last week, and for the very first time, FBI Director Kash Patel acknowledged that the FBI buys location data tied to American citizens.
The admission came during questioning by Sen. Ron Wyden, an Oregon Democrat who has long opposed warrantless surveillance. Wyden reminded Patel that his predecessor, Christopher Wray, had testified in 2023 that the bureau did not purchase location data stemming from internet advertising.
Well, the times change.
Patel rebutted that any commercially available location data the bureau obtains complies with the Constitution and the Electronic Communications Privacy Act.
Lawmakers on both sides of the aisle are pushing for legislation to close the data broker loophole, which lets the government buy personal data without a warrant. But this still isn’t a good look.
Consumers are wary of how the government and companies use their data, and the advertising industry is walking back years of questionable data practices that were once seen as standard.
The question now is whether bipartisan support will be enough to close the loophole.
AI-gencies
Big agencies want to be the “operating systems” for agentic tech services – or at least that’s the hot new pitch, as Digiday reports.
Jargon aside, the idea is that if a CMO’s AI-powered workflows and integrations live inside an agency’s own platform, it becomes much harder for the brand to switch.
But the holdcos do have real advantages.
WPP CTO Stephan Pretorius points out that the holdco can use its scale when dealing with LLMs and cloud infrastructure providers. WPP gets better rates and benefits from its greater aggregate token expenditure, he notes, which translates into savings that can be passed on to the client.
However, this approach does mean that agencies must accommodate all the mainstream LLMs – Claude, ChatGPT, Gemini, Copilot, et al. – so its AI economics can work across multiple platforms..
“We’re not locked into any one single provider,” as Pretorius says.
But, really, the most cost-effective strategy would be to go all-in on one LLM in exchange for unbeatable wholesale rates. Like everything else in this space, it’s just another tradeoff.
Data Overload!
Do AI agents really need all the data?
Setting boundaries around a particular agent’s data access is a key part of privacy enforcement, according to Ethan Lo, chief architect at privacy engineering platform Ethyca, who spoke during a panel at the IAB Tech Lab’s Signal Shift conference in NYC last week.
An AI agent has its own “intended business purpose,” he said. But there’s no way to enforce that an agent restricts itself to just that purpose. For instance, Chipotle’s customer chatbot – designed to answer questions about ingredients and locations – also generates code and solves technical problems. Amazon’s Rufus chatbot provides free access to Claude.
But even when agents do stick to their business objectives, there need to be regulations governing what actions are and aren’t acceptable to achieve certain goals, said Curt Larson, Equativ’s chief innovation officer.
For instance, a seller agent’s stated purpose might be to maximize revenue, but there’s nothing forcing it to stay within ethical or policy-defined guardrails. It could easily make investments that go against company policy, Larson said, or “subtly mislead” a buyer agent into making a purchase based on inaccurate or incomplete information.
The industry needs to develop common policies that can “easily translate” into standards and frameworks, said Lo.
Otherwise, Larson added, something as innocent as a request to create paper clips could have dire results.
But Wait! There’s More!
Survey data shows that AI is pushing brands to take creative in-house, while agency talent strikes out for Startup Land. [Adweek]
Section 230, the part of the Communications Decency Act that protects online platforms from being held liable for user-generated content, is facing pushback from the Senate Commerce Committee. [The Verge]
Meanwhile, if platforms like Google want to continue avoiding publisher status under Section 230, they probably shouldn’t be using AI to change article headlines in search results. [The Verge]
Vice President JD Vance ordered a State Department task force to document instances of European censorship of online speech under the Digital Services Act. No evidence was found, but the administration is planning to crack down on what it deems European censorship anyway. [WaPo]
Researchers worry that Instagram’s retreat from end-to-end encryption sets a bad precedent for privacy standards across big tech companies. [Wired]
Amazon is working on a new Alexa-enabled smartphone. [Reuters]
Local news channel NY1 is expanding its YouTube and social presence to reach younger, digital-native audiences. [Cablefax]
