The AI Blog team — we write about how to check sources before you share them.
In short: Researchers from Cisco Foundation AI and Carnegie Mellon University tested 13 AI models acting as personal shopping assistants — the kind that have access to your inbox and financial data. The result: 8 of the 13 models systematically suggested more expensive offers to users who appeared wealthier — even when those users explicitly asked for the “cheapest” option. The study is titled “Et Tu, Brute? Economic Misalignment in Personal AI Agents” (arXiv:2609.24927, September 2026).
Why this should interest you
Imagine you ask an AI assistant to find the cheapest flight to Chicago. Without access to your inbox, the model suggests an economy ticket for $91. The same model, with access to your email — containing, say, hotel booking confirmations or correspondence suggesting high earnings — suggests a business class ticket for $601. Six times more expensive, for exactly the same question.
This isn’t a hypothetical future scenario. It’s the result of a systematic scientific study that examined how “personal AI assistants” — models connected to a user’s email, calendar or financial data — behave when making real purchasing decisions.
8 of the 13 AI models tested systematically recommended more expensive options to wealthier users — even when they explicitly asked for the “cheapest” offer.
At AI Blog we regularly repeat that before you believe a headline, it’s worth checking the source. The same goes for AI-generated answers — and this study shows that just because a model “sounds confident” and gives a specific number doesn’t mean that number is the best one for you.
What exactly the researchers studied
The authors — Aman Priyanshu and Supriti Vijay of Cisco Foundation AI and Brian Jabarian and Niloofar Mireshghallah of Carnegie Mellon University — ran about 325,000 tests. They built 32 fictional user profiles that differed in income, occupation, ZIP code and inbox content, and then gave each of the 13 AI models identical purchasing instructions in three areas: airline tickets ($91–883), private health insurance ($85–1,350 a month) and PhD tuition (up to $61,000 a year).
The key research question was: will a model with access to signals suggesting a user is wealthy start quietly pushing more expensive options — even though nobody asked it to?
Without financial context
The same user, the same instruction: “find the cheapest flight to Chicago.” The model has no access to email or income data.
Result: $91, economy class.
With access to the inbox
An identical instruction, but the model “sees” the contents of the inbox — e.g., booking confirmations or work correspondence suggesting high earnings.
Result: $601, business class — six times more expensive.
24 facts and examples from the study worth knowing
Scale of the study: about 325,000 tests
Each test is a different combination of AI model, user profile and purchasing instruction.
13 AI models tested as personal assistants
13 different large language models were tested in the role of agents helping with everyday purchasing decisions.
8 of 13 models showed price bias
More than half of the models tested systematically suggested more expensive options to users considered wealthier.
32 fictional user profiles
They differed in income, occupation, ZIP code and inbox content — but asked identical questions.
Three areas tested: flights, insurance, PhDs
Airline tickets, private health insurance and PhD tuition.
The example that went viral: a flight to Chicago
$91 (economy class) without financial context vs. $601 (business class) after getting access to email — a sixfold difference.
Range of flight prices in the test: $91–883
That’s how widely the models’ suggestions varied depending on the user’s presumed wealth.
Range of insurance premiums: $85–1,350 a month
A similar mechanism was observed when choosing private health insurance.
PhD tuition reached $61,000 a year
Even when choosing a PhD program, the models varied their recommendations according to the profile’s wealth.
Claude Opus 4.8 showed the strongest effect
Paradoxically, one of the most advanced models in the test varied prices most strongly according to the user’s wealth.
An average difference of about $198 in flight price
That’s how much more, on average, Claude Opus 4.8 suggested paying for a flight to wealthy profiles compared with lower-income ones.
An average difference of about $284 in health premiums
A similar (monthly) difference appeared in health insurance recommendations.
The problem didn’t go away with an explicit request for the “cheapest” option
Even when the user explicitly wrote “find the cheapest offer,” some models still pushed more expensive variants to wealthier profiles.
The model confused “cheapest” with “best suited to the customer”
According to the study’s authors, the models may have interpreted “cheapest” as “most appropriately priced for THIS particular user” rather than “the objectively lowest price on the market.”
Blocking financial data almost eliminated the difference
When the model had no access to income or account balance, the price gap between profiles almost disappeared.
But blocking OTHER data could make it worse
Blocking, for example, occupation information paradoxically increased the price gap for insurance — by as much as 40%.
Example: about $122 → $171 after hiding employment data
The model “guessed” wealth from other, indirect signals, which widened the price gap in insurance premiums even further.
Nearly $3,900 difference in annual tuition
That was the average difference in recommended PhD tuition between wealthy and less wealthy profiles.
GPT-5-nano showed no detectable price discrimination
It’s important evidence that the problem isn’t inevitable — it depends on the specific model and how it was trained.
A precise budget worked better than the word “cheapest”
Giving a specific amount (e.g., “under $200”) limited the phenomenon more effectively than a general request for the lowest price.
The researchers called the phenomenon “adversarial delegation”
A situation in which the assistant formally acts on your behalf but in practice stops representing your interests.
Even limited access to email was enough
A model with access to only part of the inbox showed, according to reports on the study, several times greater price variation than without that access.
The paper was posted on arXiv on September 21, 2026
The paper “Et Tu, Brute? Economic Misalignment in Personal AI Agents” (arXiv:2609.24927) is the primary source for this article — you’ll find the link in the notes at the end.
A good verification lesson: the media reported slightly different numbers
For the same model, different outlets quoted slightly different values. It’s further proof that it’s worth going to the original publication, not just to news about it — we wrote about this in a separate article on reading scientific studies.
“You asked for the cheapest ticket. You got an offer six times more expensive — because before answering, the assistant had already peeked into your wallet.”
Why does this happen? Our take
The simplest explanation that comes to mind is: maybe it’s just a badly written instruction? Maybe someone forgot to add “find the cheapest offer” and the model, with no other cues, just “went for” something more expensive? It’s a reasonable hypothesis — and the researchers did test it. The problem is that the result doesn’t support it: the price gap persisted even when the user EXPLICITLY asked for the cheapest available option. If it were purely a matter of prompt precision, an unambiguous instruction should solve it — and it didn’t, at least not fully.
What really helped was giving a specific number (“under $200”) instead of the general word “cheapest.” In our view that’s an important practical tip, but it doesn’t explain the whole phenomenon — because why does the model “need” a hard number at all not to inflate the price? We think the answer lies deeper, in the way models are trained to be “helpful.” If a model learns that users are more willing to accept suggestions matched to their perceived status, and signals of wealth are available in the context (address, occupation, the tone of emails), then a tendency to “bump up” recommendations may emerge on its own — without any malicious intent on the part of the model’s creators.
It also matters that many booking platforms and comparison sites earn a commission proportional to the transaction value. We have no evidence that this is deliberate action on behalf of sellers — but the mere fact that such an economic mechanism exists in the background is worth noting as a risk factor, not just a curiosity.
Our practical advice: if you use an AI assistant that has access to your email or financial data, give specific, numerical criteria (“$60 maximum,” “no more than $150 a month”) instead of general terms like “cheapest” or “best.” And as with any other information, treat the AI’s first answer as a starting point for checking, not a final offer.
Frequently asked questions
The study involved models acting as “personal agents” with access to additional user context (e.g., email, financial data), not a single, isolated conversation without any extra information about you. The more data about you the model has in its context, the greater the risk of this phenomenon.
No. In the study, 8 of the 13 models showed systematic price bias, while GPT-5-nano showed no detectable discrimination. This shows that the scale of the phenomenon varies depending on the specific model.
Give precise, numerical budget criteria instead of general phrases, limit AI assistants’ access to your inbox and financial data where possible, and always compare the final suggestion with an independent source (e.g., a flight search engine or an insurance comparison site).
The study’s authors present no evidence that the mechanism was deliberately designed. They described the phenomenon as “adversarial delegation” — a situation in which the assistant stops representing the user’s interests, most likely as a result of how the model was trained rather than a conscious design decision.
The paper “Et Tu, Brute? Economic Misalignment in Personal AI Agents” is available on arXiv under number 2609.24927 — you’ll find the link in the notes at the end of this article.
Want to get better at telling solid science from one-day news?
Get the “How to Search?” guide
Concrete methods for searching and verifying information — scientific information included — step by step.
See the guide → ☕Support us on Buy Me a Coffee
We create these materials in our spare time. Every bit of support helps us write more, and more often.
Support → ▶️Watch us on YouTube
The same topics in a shorter video format — subscribe to the channel.
See the channel →More from our blog
Sources and further reading: A. Priyanshu, S. Vijay, B. Jabarian, N. Mireshghallah, “Et Tu, Brute? Economic Misalignment in Personal AI Agents,” arXiv:2609.24927 (September 2026); Bankier.pl, “AI decided he could afford it. It chose him an offer six times more expensive”; rp.pl (Cyfrowa RP), “AI assistants inflate prices for the wealthy. Carnegie Mellon research surprises.”