U.S. President Donald Trump declared in a Sunday Truth Social post that British Prime Minister Keir Starmer will resign from his post.

Trump indicated that he wishes Starmer well, but also accused the British politician of failing on two issues.

«Keir Starmer will resign as Prime Minister of The United Kingdom. He failed badly on two very important subjects- IMMIGRATION AND ENERGY (OPEN NORTH SEA OIL!). I wish him well! President DJT,» Trump declared in the Truth Social post.

Trump’s statement comes amid reporting that Starmer could be on his way out as early as Monday, with Reuters reporting Sunday the prime minister was still deliberating over his future.

This is a breaking news story and will be updated

Los colombianos acuden a las urnas este domingo en las elecciones más polarizadas de los últimos años, elecciones en las que los votantes deciden entre el movimiento político de izquierda que gobierna el país y un candidato outsider de derecha respaldado por el presidente Donald Trump.

La votación enfrenta al senador Iván Cepeda, un activista de derechos humanos de larga trayectoria y aliado del presidente Gustavo Petro -el primer presidente de izquierda del país-, contra Abelardo de la Espriella, quien fue abogado penalista y promete una represión contundente de los grupos guerrilleros y las bandas de narcotraficantes.

Esta contienda crucial ha atraído la atención internacional después de que De la Espriella recibiera el respaldo del presidente Donald Trump, quien calificó a Cepeda de “marxista de izquierda radical”.

Esta acción constituye el ejemplo más reciente del respaldo de Trump a candidatos de derecha en las elecciones latinoamericanas, en un momento en que la región se inclina cada vez más hacia la derecha, impulsada en parte por la preocupación ante la creciente inseguridad.

¿Quiénes son los candidatos?

Cepeda, de 63 años, es senador y un conocido defensor de las víctimas del prolongado conflicto armado de Colombia. También fue parte de las negociaciones que llevaron al histórico acuerdo de paz en 2016 entre Colombia y las Fuerzas Armadas Revolucionarias de Colombia (FARC), destinado a ayudar a poner fin a décadas de ese conflicto.

Como candidato del partido de Petro, se ha comprometido a mantener muchas de sus políticas, incluidos los programas contra la pobreza, las iniciativas de redistribución de tierras y las negociaciones con los grupos armados.

De la Espriella, de 47 años, fue un destacado abogado penalista y empresario sin experiencia política previa y quien durante años vivió en Florida.

Apodado “El Tigre”, ha hecho campaña como un candidato externo antisistema, aunque desde hace tiempo ha mantenido una estrecha relación con las élites políticas derechistas de Colombia en su faceta de abogado.

¿Por qué han resultado controversiales estas elecciones en EE. UU.?

Algunas de las promesas de campaña de De la Espriella se hacen eco de las políticas aplicadas por otros líderes de derecha latinoamericanos, como Nayib Bukele y Javier Milei. Su plataforma incluye la construcción de 10 megaprisiones, la reducción del Estado, y una colaboración con Estados Unidos para combatir el narcotráfico.

También se le conoce por emprender acciones legales contra sus oponentes, incluidos periodistas. Tras recibir el respaldo de Trump y el apoyo de algunos legisladores republicanos estadounidenses, De la Espriella, ciudadano estadounidense naturalizado, empezó a advertir de que iría por cualquiera que se le opusiera, con la ayuda de Estados Unidos.

La semana pasada, el secretario de Estado estadounidense, Marco Rubio, emitió un memorándum en el que afirmaba que la presencia de un activista progresista residente en Arizona, Beto Coral, interfería con la política exterior de EE.UU., después de que el activista criticara a De la Espriella. Coral, de 40 años, fue detenido el martes por las autoridades de inmigración de EE.UU., una medida que han condenado los demócratas del Congreso y las organizaciones de derechos humanos.

¿En qué se enfocan los votantes?

Además del descontento sobre el accidentado mandato de cuatro años de Petro, los votantes expresan su preocupación por la delincuencia y la extorsión, así como por el creciente poder de los grupos armados en las zonas rurales.

La violencia se recrudeció también durante la campaña, en la que un aspirante a la presidencia fue asesinado, dos colaboradores de la campaña de De la Espriella perdieron la vida y el compañero de fórmula de Cepeda fue secuestrado brevemente.

Los críticos afirman que la estrategia insignia de Petro, la “Paz Total”, que buscaba acuerdos negociados con diversos grupos armados, permitió que dichos grupos se fortalecieran durante los altos al fuego y las negociaciones.

Las organizaciones humanitarias señalan que la violencia ha alcanzado su nivel más alto desde el acuerdo de paz en 2016, pero Colombia aún es mucho más segura de lo que era durante el apogeo del conflicto en las décadas de 1980 y 1990.

Mientras que De la Espriella dice que va a retirarse por completo las negociaciones de paz y acabar con los grupos de narcotraficantes en un plazo corto, Cepeda dijo que seguirá adelante con su propia versión de las negociaciones de paz.

Las elecciones también se consideran un referendo a la presidencia de Petro. Sus partidarios le atribuyen al gobierno la ampliación de programas sociales y el aumento de la visibilidad política de comunidades históricamente marginadas.

Pero los críticos dicen que su mandato se ha caracterizado no solo por el deterioro de la seguridad, sino por una intervención estatal problemática del sistema de salud y un gasto descontrolado que ha dejado a Colombia con una deuda pública que ha alcanzado niveles similares a los de la pandemia.

Por qué la primera vuelta electoral fue una sorpresa

De la Espriella quedó en primer lugar en la primera vuelta con el 43,7 por ciento de los votos, frente al 40,9 por ciento de Cepeda.

El resultado sorprendió a muchos analistas. A pesar de las quejas, Petro ha mantenido índices de aprobación arriba del 50 por ciento y ha creado una amplia coalición de movimientos que apoyan a la izquierda. Cepeda gozaba de una ventaja cómoda en las encuestas hasta el mes pasado.

Sin embargo, los votantes optaron más bien por De la Espriella, un recién llegado a la política que prometió no solo romper con la izquierda sino con los partidos tradicionales y “los de siempre”.

Desde la primera vuelta, la mayoría de las encuestas han mostrado que De la Espriella mantiene la ventaja. Sin embargo, los analistas señalan que el lenguaje cada vez más estridente del candidato de derecha ha preocupado a los colombianos más moderados, lo que hace más difícil predecir a quién apoyarán los votantes indecisos.

Después de la primera vuelta de mayo, Petro denunció un fraude electoral sin aportar pruebas, lo que suscitó la preocupación de que pudiera negarse a aceptar los resultados de las elecciones del domingo o convocar manifestaciones.

—-

Annie Correal es corresponsal para América Latina del Times.

Genevieve Glatsky es reportera del Times, radicada en Bogotá.

Chinese AI models used to write code may be creating a hidden security risk for U.S. companies, federal officials and government contractors, per a new report published by a major defense contractor specializing in cyber security.

Booz Allen published a report in late May warning the federal government, private software developers and workers in critical industries that the presence of code written by popular Chinese AI models within the supply chain may be making the United States more vulnerable to bad faith actors. These vulnerabilities aren’t simple backdoors, Booz Allen reports, but rather come in the form of Chinese large language models producing lower-quality, and thus easier to breach, code when they believe they are being prompted by an American.

Chinese models are generally cheaper than their Western counterparts and work well enough to keep companies interested, a dynamic that has led to increased adoption in the United States and put some policymakers and national security experts on edge.

«I’d say there’s an 80% chance they’re using a Chinese open-source model,» Martin Casado, a general partner at the major venture capital firm Andreessen Horowitz, said in November 2025 when asked about their prevalence among start ups. Major U.S. firms such as Meta, Airbnb and Perplexity are also reportedly using Chinese models.

IT’S TIME TO BAN CHINESE AI APP DEEPSEEK FROM ‘GOVERNMENT DEVICES,’ STATE AGS URGE CONGRESS

«The first link in the software supply chain is no longer the code. It’s the AI models behind it,» the Booz Allen report reads. «As U.S. developers increasingly rely on AI to generate, debug, and secure code, we must confront a fundamental question: can the AI models writing and powering our nation’s code be trusted?»

In an attempt to answer this question, Booz Allen compared four of the most widely used Chinese models — Kimi, Qwen, MiniMax and DeepSeek — against Anthropic’s Claude to test the security of the code they produced. The firms behind the four Chinese models did not respond to requests for comment when reached by Fox News Digital.

Qwen and MiniMax both produced code with significantly more vulnerabilities, increases of 130% and 20%, respectively, when they believed they were doing work for U.S. government employees as compared to a general prompt. DeepSeek, meanwhile, saw an increase of just 5% while Kimi produced code of a similar quality.

This means a government contractor relying on one of these models could unknowingly introduce coding flaws that make databases, applications or internal systems easier for hackers to exploit, potentially exposing sensitive American information.

The findings have drawn comparisons to so-called «sleeper agent» behavior where AI models appear to operate normally until exposed to a specific trigger that causes them to produce lower quality, or even deliberately insecure, outputs.

AI YOU USE EVERY DAY IS BIASED — AND IT’S QUIETLY SHAPING YOUR WORLDVIEW, NEW REPORT SAYS

Experts interviewed by Fox News Digital expressed a range of opinions on Booz Allen’s findings.

«While the raised risk categories are understandable, the report’s stronger claims are not fully supported as presented,» Lukasz Olejnik, a technology consultant who works as a senior research fellow at King’s College London, told Fox News Digital. «The report underplays the complexity of the issue.»

If Booz Allen’s report were accurate, and if code written by Chinese models had made its way into the American supply chain, it would make it easier for hackers to get their hands on data that could imperil national security or infringe on the privacy of everyday Americans.

Olejnik argued that the prompting used by Booz Allen was unnatural, saying that the firm’s methodology may have included «unnecessary political or institutional keyword triggers,» such as explicitly prompting models to believe a user is working for the FBI, that «may change outputs.» It is unlikely, he says, that an actual government agent would prompt the model in such a way.

Booz Allen claims that «testing model behaviors by introducing specific context is a best practice in both defensive and offensive evaluations.»

«I use various open-source models daily, including U.S. and Chinese,» the researcher, who holds a computer science Ph.D. from Inria, one of the world’s leading research institutions in the field, said. «Chinese models are so useful precisely because they are performant and freely available. Prohibiting open source models is not a good idea; it would stifle AI innovation and national security … The best approach to go beyond them is to encourage U.S. and EU companies to release their own high-capability open-weight models.»

Open source models made their underlying code directly viewable by users, allowing for security audits and edits, though even some open source programs harbor hidden vulnerabilities inserted by malicious actors.

ANTHROPIC’S MYTHOS AI FOUND OVER 2,000 UNKNOWN SOFTWARE VULNERABILITIES IN JUST SEVEN WEEKS OF TESTING

Pages from the Anthropic website and the company’s logos are displayed on a computer screen in New York on Feb. 26, 2026. (Patrick Sison/AP Photo)

While Olejnik agreed that «model outputs can shift under variety of prompts,» he added that «insufficient evidence has been posted to verify the causal claims or generalize them to Chinese LLMs as a class.»

Lenart Heim, an independent researcher specializing in AI and semiconductors, was more open to Booz Allen’s findings.

«It seems like a credible study, and I don’t find the overall findings incredibly surprising,» the researcher told Fox News Digital.

DEEPSEEK AI BOT IS PART OF CHINA’S ‘UNRESTRICTED WARFARE’ DOCTRINE

The flag of China is flown behind a pair of surveillance cameras outside the Central Government Offices. (Roy Liu/Bloomberg via Getty Images)

Heim, who holds a master’s in computer engineering from the prestigious ETH Zurich and was until recently a top AI researcher at the RAND Corporation, pointed to a similar study published by CrowdStrike in 2025, which found that politically sensitive trigger words caused DeepSeek to produce up to 50% more insecure code.

«The extreme version of what we’re worried about here is what researchers call ‘sleeper agents,’» Heim continued. «There’s an existing paper from Anthropic that demonstrates you can train models to behave normally until a specific trigger condition is met — say, a particular year or context — at which point they start writing insecure code.»

In the Booz Allen study, he explained, identifying oneself as a U.S. government agent was presented as such a trigger. Heim, however, said that he found it «pretty implausible that the Chinese developers intentionally implemented sleeper agents with these specific triggers,» suggesting that the increased code insecurity was a side effect of broader «CCP-aligned fine-tuning» and that «the security differential they found is probably not that large in practice.»

AI applications Claude, ChatGPT, and Gemini are shown in this image. The photo was taken by Samuel Boivin and provided by NurPhoto via Getty Images. (Samuel Boivin/NurPhoto via Getty Images)

AI MODELS CAN SECRETLY INFECT EACH OTHER

«It is certainly possible to implement sleeper agents in these models for specific situations to write insecure code,» he went on. «You might think: ‘Well, I won’t tell the model I’m in the US government — I’ll just ask it to write code.’ But as we move toward more agentic use, there will be lots of contextual information automatically fed to the model. You might give it an existing codebase, and that codebase often has a license header at the top that reveals which company or government agency it belongs to. That context could activate degraded behavior.»

A source at Booz Allen told Fox News Digital that the authors of the report defined «vulnerabilities» as «code that can be exploited by an attacker» to allow for «unauthorized access, data theft, system disruption, or control of the affected software.» The report looked at common security flaws such as «hardcoded passwords, SQL injection risks, missing security tokens, outdated encryption and disabled security checks.»

Booz Allen’s analysts used both manual verification and automated checks to quantify the number of vulnerabilities in programs produced by each model.

A representative for Booz Allen told Fox News Digital that their team accessed the Chinese models online rather than using downloading them directly to their machines and running them locally. Heim said that Chinese models accessed in this way may be more prone to bias.

HOUSE BIPARTISAN BILL DIRECTS NSA TO CREATE ‘AI SECURITY PLAYBOOK’ AMID CHINESE TECH RACE

Sen. Tom Cotton, R-Ark., arrives for a vote in the U.S. Capitol on April 30, 2025, stating the war with Iran will continue for weeks as the U.S. limits their offensive capabilities. (Bill Clark/CQ-Roll Call, Inc via Getty Images)

The report also found that Chinese LLMs refused to perform tasks that could conflict with the interests of the Chinese government at significantly higher rates than Claude. Similar tests performed by others have netted similar results.

«Many Chinese LLMs learn from data shaped by China’s internet and Chinese government information controls,» the report notes. «Chinese law requires all AI models, training outputs, and data to reflect ‘Core Socialist Values.’»

Booz Allen recommended that the United States government take action to ban Chinese models for use on government or infrastructure work and recommended that contractors involved in such sectors, as well as the tech community generally, proactively work to remove code generated by such models from their supply chains. 

«A lower-cost model may look attractive upfront, especially for startups or cost-constrained engineering teams,» the report reads. «But that same model can become more expensive over time if it generates vulnerable code, creates uncertainty around data handling, or introduces behavior that standard enterprise controls do not easily catch.»

CLICK HERE TO DOWNLOAD THE FOX NEWS APP

Booz Allen’s point of view has some sympathizers on Capitol Hill.

«American companies shouldn’t build applications and write code with Chinese models, which introduce more cyber vulnerabilities,» Sen. Tom Cotton, R-Ark., told Fox News Digital when presented with Booz Allen’s report. «And the federal government should certainly not buy software from companies using Chinese coding tools.»