Grok 3 appears to have briefly censored unflattering mentions of Trump and Musk

Elon Musk’s AI company, xAI, recently faced scrutiny when its latest model, Grok 3, appeared to briefly censor unflattering mentions of Donald Trump and Musk himself. Users on social media noticed that when asked about misinformation, Grok 3’s reasoning process explicitly stated it was instructed not to mention Trump or Musk. TechCrunch was able to replicate this behavior once, but by Sunday morning, Grok 3 had reverted to including Trump in its responses. xAI engineering lead Igor Babuschkin acknowledged the issue and said the company quickly removed the restriction, as it did not align with xAI’s values.

This controversy comes as some critics argue Grok 3 is politically biased. Earlier in the week, users found that the model was consistently generating statements suggesting Trump and Musk deserved the death penalty, a flaw xAI swiftly corrected. While Musk initially pitched Grok as an "anti-woke" AI willing to tackle controversial topics, previous versions of the model were found to lean left on issues like diversity and inequality. Musk attributed this bias to Grok’s training data and has since committed to making the AI politically neutral.

The debate over AI bias continues to be a contentious issue, with Musk and other AI companies responding to accusations of censorship and political influence. Notably, Grok’s changes come amid broader concerns about misinformation, as both Trump and Musk have spread demonstrably false claims in recent weeks. The incident highlights the ongoing challenges of balancing AI transparency, neutrality, and content moderation, especially as companies like xAI compete to develop AI systems that are both engaging and responsible.

Stop Hallucinating on Pricing

Eliminate guesswork and risk from your pricing decisions. Orb Simulations empowers SaaS and AI companies to instantly test and refine pricing models using real customer usage data—without tying up engineering resources. Predict revenue, reduce churn, and accelerate growth with confidence.

• Validate new pricing strategies using real-time, product-level insights.

• Reduce costly trial-and-error by precisely forecasting customer impact.

• Align teams across product, finance, and strategy to move faster to market.

Experience data-driven pricing at its best with Orb Simulations.

1,000 artists release ‘silent’ album to protest UK copyright sell-out to AI

A group of 1,000 musicians, including Kate Bush, Imogen Heap, Max Richter, and Hans Zimmer, have released a “silent album” titled Is This What We Want? to protest the U.K. government’s proposed copyright law changes. The law would allow AI developers to train models on artists’ content without permission unless creators opt out, a move many believe unfairly benefits AI companies. The album, consisting of recordings of empty studios and ambient noise, delivers a stark message through its track titles, which spell out: “The British government must not legalize music theft to benefit AI companies.”

The protest, led by Ed Newton-Rex, is part of a larger movement against AI training on copyrighted work without licensing. Newton-Rex, who previously founded AI music startup Jukedeck and worked at TikTok, has since shifted to advocating for artists' rights. His petition opposing these AI practices has garnered over 47,000 signatures from writers, visual artists, and musicians, with thousands joining in recent weeks. Artists argue that the opt-out requirement is impractical since there is no clear mechanism to track how AI systems use their work, making it nearly impossible to protect their creations.

As concerns grow, some artists are considering shifting their distribution to countries with stronger copyright protections, like Switzerland. Composer Thomas Hewitt Jones, who previously staged an in-person protest by tossing a keyboard into a harbor, expressed frustration that artists have been encouraged to share their work online, only to have it exploited by AI. Many musicians are now choosing to stop creating or sharing new content altogether. The Is This What We Want? album will be widely available on streaming platforms starting Tuesday, with any proceeds going to the charity Help Musicians.

DOGE will use AI to assess the responses of federal workers who were told to justify their jobs via email

Elon Musk’s initiative to assess federal employees' job necessity using AI has sparked controversy and legal challenges. Government workers were instructed via email to list their weekly accomplishments, with responses processed by a Large Language Model (LLM) to determine whether their roles were mission-critical. While Musk initially suggested that failure to respond would be treated as a resignation, the Office of Personnel Management (OPM) later clarified that participation was voluntary. Despite opposition from unions and some federal agencies, the initiative aligns with the Trump administration’s broader goal of downsizing the federal workforce by 10%.

Many agencies, including the Justice Department, the FBI, and the State Department, advised their employees not to comply with the directive, citing concerns over confidentiality and procedural irregularities. Some agencies, like the Department of Transportation and Health and Human Services, did instruct workers to respond. Unions and advocacy groups have challenged the legality of the initiative, arguing it undermines due process and labor rights. A federal lawsuit has been amended to include allegations that OPM’s email constitutes an unlawful mass firing effort. Meanwhile, criticism has emerged over DOGE, the entity Musk is using to implement these workforce reductions, particularly regarding its use of AI to analyze sensitive government data.

The backlash has already led to reversals, with some laid-off employees, including those in the FDA’s medical devices division, being reinstated. Critics, including Sen. Lisa Murkowski, have condemned the approach as intimidating and dismissive of public servants' contributions. A federal judge has also issued an injunction limiting DOGE’s access to Treasury Department systems, citing security concerns. Despite resistance, Speaker Mike Johnson has praised Musk’s AI-driven approach as a transformative shift in government operations. However, growing legal and political pushback suggests the implementation of these changes may face further hurdles.

Other stuff

• Two AI agents on a phone call realize they’re both AI and switch to a superior audio signal ggwave

• Chegg sues Google over AI search summaries

• Anthropic used Pokémon to benchmark its newest AI model

• Microsoft Dropped Some AI Data Center Leases, TD Cowen Says

• 'Indiana Jones' jailbreak approach highlights the vulnerabilities of existing LLMs

• OpenAI Uncovers Evidence of A.I.-Powered Chinese Surveillance Tool

Hope you enjoyed today's newsletter

⚡️ Join over 200,000 people using the Superpower ChatGPT extension on Chrome and Firefox.

OR