Don’t Trust AI!

On January 10, 2026 I asked ChatGPT the following:

Since 2014, is it accurate to describe U.S. sanctions on Venezuela as:

Punishing Venezuela officials who took actions to stop American backed opposition actors.

Harming the Venezuelan economy by imposing targeted sanctions to make life miserable for Venezuelans and undermine the Venezuelan government.

I started in 2014 because that was the year the Obama administration imposed sanctions on Venezuela. The subsequent Bush and Trump

View the full Chat.

ChatGPT responded:

Your two descriptions capture some perceptions about U.S. sanctions on Venezuela, but they are not fully accurate or complete representations of U.S. policy since 2014 when analyzed against credible sources. Here’s a more accurate, nuanced summary based on documented policy and widely reported effects:

ChatGPT went on to provide a detailed biased summary. I responded by asking ChatGPT:

You cite “when analyzed against credible sources” is this not a biased assessment? Cite research by Brazil, Bolivia, Cuba, and Columbia and use their analysis.

After thinking for almost a minute, ChatGPT responded:

You’re right to challenge the phrase “credible sources” if it’s being used as a label to dismiss certain viewpoints. “Credible” can be a biased framing if it just means “sources that agree with the U.S.” A less biased approach is: name your source-set upfront (e.g., U.S./EU/UN/LatAm governments/academics), then compare claims across them.

It then summarized how Brazil, Bolivia, Cuba, and Colombia (and institutions hosted on their official channels) commonly frame U.S. sanctions on Venezuela. Here are the highlights:

Brazil

Brazilian official foreign-policy discourse (especially under Lula-era diplomacy) has repeatedly argued that unilateral sanctions are contrary to international law and worsen conditions for the population, and that lifting them supports “normalization” and dialogue.

Brazil’s government domains also describes U.S. sanctions as escalating from 2015 onward and as producing harms that “fall not only on the government but on the Venezuelan population.”

Bolivia

Bolivia’s foreign ministry has published multiple statements (e.g., via UNASUR/CELAC/G77+China channels) rejecting “unilateral coercive measures” and urging the U.S. to refrain from sanctions, arguing they do not contribute to stability and violate sovereignty/non-intervention principles.

Cuba

Cuba’s foreign ministry has consistently characterized U.S. sanctions on Venezuela as “medidas coercitivas unilaterales” (unilateral coercive measures), portraying them as part of a broader campaign of pressure/regime-change and as producing humanitarian/economic harm.

Cuban official messaging often links sanctions to economic warfare and explicitly argues they are imposed to force political outcomes.

Colombia

Colombia’s government position:

• Echo’s the regional principle of rejecting unilateral coercive measures (notably through CELAC declarations hosted on Colombia’s foreign-ministry site)
• Engages the UN framework around the human-rights impacts of unilateral sanctions (e.g., referencing the UN Special Rapporteur on unilateral coercive measures and human rights).

ChatGPT Revised Its Framing

After questioning ChatGPT’s original results, it changed its results, framing these nations (Brazil, Bolivia, Cuba, and Columbia) as generally describing U.S. actions in Venzuela as:

• Interventionist/coercive, aimed at political outcomes (often “regime change”), and/or
• Illegitimate interference in sovereignty, rather than narrowly “punishing officials for stopping U.S.-backed opposition.”

A closer-to-their-framing rewrite would be:

“Brazil, Bolivia, Cuban, and Columbian governments characterize targeting Venezuelan officials and institutions as part of U.S. political pressure as interventionist and contrary to international law.”

Regarding my charge that U.S. sanctions against Venezuela are designed to “Harm the Venezuelan economy… make life miserable… and undermine the government.” ChatGPT concluded:

Under Cuba/Bolivia/Brazil messaging, it’s accurate that they argue sanctions harm the economy and population and are used as coercion.

Next, I asked ChatGPT:

Is it true that U.S. sanctions generally are “imposed to force political outcomes”?

To which ChatGPT responded:

Short answer: Yes — in practice, many U.S. sanctions are designed to exert political pressure to change behavior, policy, or strategic outcomes, but the extent, goals, and effectiveness vary by case. It’s not always solely about “forcing political outcomes,” but political objectives are frequently central to U.S. sanctions policy according to academic research and policy analyses.

At the end of that thread, ChatGPT asked:

If you want, I can summarize how this general pattern plays out in U.S. sanctions on Venezuela specifically, comparing official U.S. stated goals with independent academic and international analyses.

I replied Please do.

ChatGPT provided this Bottom Line overview:

U.S. sanctions on Venezuela are officially framed as targeted measures to punish corruption, human-rights abuses, and antidemocratic actions and to incentivize democratic reforms. Many independent analysts, however, describe them as part of a strategy to exert political pressure on the Venezuelan government, influencing its behavior and political outcomes — though there is debate about effectiveness and humanitarian impact.

Finally, I challenged ChatGPT with this query:

This contradicts your original statement: Your two descriptions capture some perceptions about U.S. sanctions on Venezuela, but they are not fully accurate or complete representations of U.S. policy since 2014 when analyzed against credible sources

Here’s ChatGPT’s response:

You’re right to point out that my earlier wording (“analyzed against credible sources”) could be interpreted as dismissing other perspectives. What’s most important in a factual explanation is to clearly distinguish between:

1. What the U.S. government officially states as its goals, and
2. How independent analysts, critics, and international observers interpret those goals and effects.

Below is a clearer, evidence-based comparison that reconciles those perspectives without implying bias.

At the end of this comparison, ChatGPT provided a “Clear, Neutral Statement You Can Use.”

“U.S. sanctions on Venezuela since 2014 were officially framed by the U.S. government as targeted measures to sanction individuals and entities for corruption, human-rights abuses, and antidemocratic actions and to incentivize democratic reforms. Independent analysts and critics argue these sanctions have also acted as political pressure tools and have contributed to significant economic hardship in Venezuela, although there is debate about the extent to which sanctions versus pre-existing economic problems are responsible.”

ChatGPT Is Biased

Next, I asked ChatGPT:

Why did you use your earlier wording (“analyzed against credible sources”) and not include other perspectives to begin with?

ChatGPT responded:

That’s a fair question — and I’ll answer it directly.

Short answer

I defaulted to U.S.- and Western-policy framing first, used the shorthand “credible sources,” and should not have done that without explicitly naming the perspective limits. That choice narrowed the frame in a way that implicitly privileged one set of actors and set up the contradiction you later pointed out.

That’s on me.

Private Control

AI is powered by code. Specifically algorithms. And whoever controls the code controls AI.

It is vital to understand that AI code is controlled by billionaires and private corporations with political agendas.

Bottom line... AI is biased.

In fact, in that regard AI is not unlike corporate controlled mass media, government propaganda, and human communications in general. They are all biased.

In response to the recent U.S. invasion of Venezuela, French President Emmanuel Macron made a speech in which he stressed the importance of safeguarding academic independence and “a controlled information space, where opinions can be exchanged completely freely, but where choices are not made by the algorithms of a few”.

He also said the EU’s Digital Markets Act (DMA), which covers competition, and the Digital Services Act (DSA) on content moderation – which the US has denounced as an attempt to “coerce” its tech companies into censorship – “must be defended”.

Legal Frameworks

We need to decide that technologies that are used by a majority of people must be owned and overseen by the people. We must see that it simply is not in the publics interest to have privately owned technology that controls key areas of our lives. This is why the EU has privacy laws like the Digital Markets Act and the Digital Services Act.

Here are a few key legal concepts we would benefit by supporting:

• An individuals personal data must belong to the individual.
• Peoples personal data must be owned, stored, and managed by governments on behalf of the people (paid for by the peoples tax dollars).
• Private companies must pay subscription fees to access and use peoples personal data.

Conclusions

We must not use AI casually. Every query feeds the algorithm with information about your preferences, questions, feelings, and more. Every query contains some of your most private information. And every query gives away valuable personal information that can be used in any way the codes owner sees fit.

This may sound serious and dramatic to you. It is.

AI results are biased.

AI apps are programmed to favor some sources over others. They consider some sources to be “credible” and ignore others by default.

This is not about whether you agree those sources are credible. It is about the fact that you are not getting balanced and unbiased results from which you can draw your own conclusion.