Brian Merchant, reporting for Blood in the Machine, notes that individuals throughout the United States are tearing down and damaging Flock surveillance cameras, fueled by escalating public outrage that the license plate readers assist U.S. immigration officials and deportations.

Flock is a surveillance startup based in Atlanta that was valued at $7.5 billion last year and produces license plate readers. It has been criticized for granting federal authorities access to its extensive network of nationwide license plate readers and databases at a time when U.S. Immigration and Customs Enforcement increasingly depends on data for community raids as part of the Trump administration’s immigration enforcement efforts.

Flock cameras enable authorities to monitor individuals’ movements by capturing images of their license plates from a multitude of cameras spread across the United States. Flock asserts that it does not share data directly with ICE, but reports indicate that local police have provided federal authorities with their own access to Flock’s cameras and databases.

While various communities are urging their municipalities to terminate contracts with Flock, others are taking action independently.

Merchant highlights occurrences of damage to Flock cameras in La Mesa, California, mere weeks after the city council agreed to extend the deployment of Flock cameras in the city, notwithstanding a decisive majority of participants advocating for their removal. A local report mentioned significant opposition to the surveillance technology, with residents expressing privacy concerns.

Instances of vandalism have been reported from California and Connecticut to Illinois and Virginia. In Oregon, six license plate-reading cameras mounted on poles were severed, and at least one was spray-painted. A message left at the base of the cut poles read, “Hahaha get wrecked ya surveilling fucks,” according to Merchant.

As per DeFlock, a project focused on mapping license plate readers, nearly 80,000 cameras exist across the United States. Numerous cities have thus far rejected Flock’s cameras, and some police departments have blocked federal authorities from utilizing their resources.

A spokesperson from Flock did not specify, when contacted by TechCrunch, whether the company monitors how many cameras have been damaged since their installation.

OpenAI is enhancing collaborations with four key consulting firms as the AI organization aims to expand its enterprise sector in 2026.

On Monday, OpenAI revealed the “Frontier Alliances,” indicating that the AI laboratory is open to exploring various strategies to encourage enterprises to significantly adopt its technology. This alliance features long-term collaborations between OpenAI and four leading consulting companies: Boston Consulting Group (BCG), McKinsey, Accenture, and Capgemini, aimed at promoting its enterprise offerings.

The Forward Deployed Engineering team at OpenAI will partner with these consulting leaders to assist them in integrating OpenAI’s enterprise-centric technologies, such as OpenAI Frontier, into their clients’ technology infrastructures.

OpenAI introduced OpenAI Frontier in early February. This no-code open software enables users to create, deploy, and manage AI agents that are built on OpenAI’s AI models and other frameworks.

In its most recent announcement, OpenAI contends that consulting firms are the ideal channels for onboarding enterprises.

“AI alone does not drive transformation. It must be linked to strategy, built into redesigned processes, and adopted at scale with aligned incentives and culture to deliver sustained outcomes,” BCG CEO Christoph Schweizer stated in OpenAI’s blog. “Our extended partnership integrates OpenAI’s Frontier platform with BCG’s extensive industry, functional, and technological expertise along with BCG X’s capabilities for building and scaling to generate measurable impact with safeguards from the outset.”

So far, the pace of enterprise adoption of AI has been relatively sluggish as these organizations face challenges in attaining a meaningful return on investment from their AI initiatives.

OpenAI’s collaborative approach is logical and extends beyond merely encouraging enterprises to incorporate AI into their current workflows. This initiative prioritizes consultants in convincing organizations to adjust their strategies and operations to integrate OpenAI’s tools where applicable.

It is noteworthy that OpenAI’s competitor Anthropic has also signed agreements with major consulting firms, including Deloitte and Accenture, in recent months.

Company CFO Sarah Friar expressed in a blog post in January that the enterprise segment is a significant focus for OpenAI in 2026. Furthermore, OpenAI has secured substantial enterprise AI agreements with Snowflake and ServiceNow so far this year, alongside appointing Barret Zoph to head the company’s enterprise sales division in January.

The difficulty in managing a deep learning model is often deciphering why it behaves the way it does: be it xAI’s continuous attempts to refine Grok’s peculiar politics, ChatGPT’s issues with flattery, or common hallucinations, navigating through a neural network with billions of parameters is challenging.

Guide Labs, a startup based in San Francisco and led by CEO Julius Adebayo and chief science officer Aya Abdelsalam Ismail, is presenting a solution to this dilemma today. On Monday, the firm made public an 8-billion-parameter LLM, Steerling-8B, trained using a fresh architecture aimed at making its actions straightforwardly interpretable: Each token generated by the model can be traced back to its roots within the training data of the LLM.

This can range from simply identifying the reference materials for facts referenced by the model, to more intricate tasks such as grasping the model’s concept of humor or gender.

“If I possess a trillion ways to encode gender, and I utilize 1 billion of those trillion options, it’s essential to ensure that you discover all those 1 billion elements I’ve encoded, and then you need to be capable of turning them on or off reliably,” Adebayo conveyed to TechCrunch. “Current models can accomplish this, but it’s highly unstable… It’s essentially one of the ultimate questions.”

Adebayo commenced this research while pursuing his PhD at MIT, co-authoring a well-referenced paper in 2018 that demonstrated that existing techniques for comprehending deep learning models were unreliable. This endeavor eventually resulted in a novel method of constructing LLMs: Developers embed a concept layer within the model that categorizes data into traceable segments. Although this necessitates greater upfront data labeling, they leveraged other AI models to assist, allowing them to train this model as their most significant proof of concept thus far.

“The type of interpretability that individuals generally perform is… neuroscience on a model, and we turn that around,” Adebayo stated. “What we do is actually construct the model from the foundations up so that you don’t require neuroscience.”

One concern regarding this methodology is that it may remove some emergent behaviors that render LLMs fascinating: Their capability to generalize in innovative ways about concepts not previously encountered. Adebayo asserts that this still occurs in his company’s model: His team monitors what they term “discovered concepts” that the model unearthed autonomously, such as quantum computing.

Adebayo contends that this interpretable architecture will be essential for everyone. For consumer-oriented LLMs, these strategies should empower model creators to undertake actions such as blocking copyrighted materials or improving output control concerning topics like violence or substance abuse. Regulated sectors will necessitate more manageable LLMs — for instance, in finance — where a model assessing loan candidates must focus on aspects like financial histories while disregarding race. Additionally, there is a demand for interpretability within scientific endeavors, another domain where Guide Labs has innovated technology. Protein folding has seen substantial success with deep learning models, yet researchers require deeper understanding regarding why their software identifies promising combinations.

“This model exemplifies that training interpretable models is no longer merely a scientific inquiry; it has now become an engineering challenge,” Adebayo remarked. “We have deciphered the science, and we can scale them. There’s no reason this kind of model shouldn’t achieve performance on par with frontier-level models,” which contain significantly more parameters.

Guide Labs states that Steerling-8B can attain 90% of the capabilities of current models while utilizing less training data, due to its innovative architecture. The next phase for the company, which originated from Y Combinator and gathered a $9 million seed funding from Initialized Capital in November 2024, is to construct a larger model and start providing users with API and agentic access.

“The current method of training models is quite primitive, thus making the democratization of inherent interpretability a long-term benefit for our role within humanity,” Adebayo conveyed to TechCrunch. “As we pursue models that will possess superintelligence, you don’t want an entity making choices on your behalf that remains somewhat enigmatic to you.”

An AI-driven news application named Particle, developed by ex-Twitter engineers, is now capable of keeping up with breaking news from both podcasts and online articles.

Just before its latest Android launch, Particle unveiled a functionality known as Podcast Clips, which finds the most captivating and pertinent moments from a variety of podcasts, incorporating those clips next to corresponding news articles in its feed.

This means instead of wading through an entire podcast to catch 45 seconds of engaging commentary, you can listen to the clip while perusing the news on Particle. Alternatively, you can opt to read the transcript of the clip, with the words highlighted in real-time as they are spoken.

“We basically capture that for every news item — if there’s a relevant podcast discussing it, we’ve gathered those segments,” stated Particle CEO Sara Beykpour, who was formerly Twitter’s Senior Director of Product Management, in an interview with TechCrunch. “It provides an engaging way to get insights on what people are discussing about a story while you’re reading it.”

This enhancement recognizes a transformation in the news landscape that has been developing over the years. More individuals are sourcing their news from podcasts, considering them as credible outlets, and the format is becoming a go-to for urgent news and significant statements from public figures.

Tech executives, in particular, are reaching out to accommodating podcast hosts to share their narratives instead of dealing with traditional media channels, as reported by Bloomberg in 2024.

This elevates the importance of podcasts if one aims to stay updated with the news.

Beykpour mentions that Particle employs embedding models to determine when podcasts are related to specific news topics. These models are supplied by the same firms that create LLM models, though they do not pertain to generative AI technologies, she clarifies.

“We utilize vector embeddings to ascertain these segments of podcasts relate to varied stories,” notes Beykpour. “A single podcast may touch on 10 or 20 stories, so we leverage AI for understanding that. Additionally, we employ AI for some aspects of the clipping process, including determining when to commence and conclude a clip.”

The organization utilizes technology from ElevenLabs for transcription. However, some of the technology that dictates where to clip the audio is integral to Particle’s proprietary methods.

The initiative to utilize podcasts for gaining insights into news-related commentary is also being examined closely by newsrooms today. As reported by Nieman Lab this month, The New York Times has been employing a tailored AI tool that uses LLMs to transcribe and summarize fresh episodes of various right-leaning and conservative podcasts to better grasp what influencers from that sphere are expressing about current events.

Particle’s Podcast Clips function is not exclusively linked to news articles. Since the app recognizes various entities — including individuals, locations, or objects — users can visit the page of notable persons, like OpenAI CEO Sam Altman, to review all his podcast appearances presented in a feed format.

Particle has also been working on additional features. The company has initiated its first monetization effort with Particle+, a voluntary subscription priced at $2.99/month (or $29.99/year) that grants access to premium functionalities. These include the option to use natural language to obtain news summaries in a preferred style; select from various voices during the personalized audio feed; “Listen to the News”; unlimited crossword puzzles; the ability to ask private questions to its AI chatbot; and more.

The Android version also introduces several other noteworthy changes. The browsing tab now features timely topics, such as the 2026 Winter Olympics, in addition to standard sections like politics, technology, or entertainment. Furthermore, when clicking on an entity, users will observe a fresh page presenting definitions, stories, articles, related entities, and associated topics.

Particle has not disclosed information regarding user engagement stats or conversion rates, but Beykpour did mention the app’s global audience prior to the Android launch. Weekly, 55% of Particle’s users are located outside the U.S., with India (15%) being its largest market following the U.S.

Following a successful test of its AI-driven “Prompted Playlist” feature in New Zealand and a recent launch in the U.S. and Canada, Spotify revealed on Monday that it will be introducing this tool to Premium users in the U.K., Ireland, Australia, and Sweden.

The Prompted Playlist feature enables users to generate personalized playlists by simply articulating their musical preferences in their own words. Rather than searching for specific songs or artists, users can outline the mood, circumstance, or inspiration they seek, while Spotify handles the rest.

To utilize the feature, users click “Create” and then choose “Prompted Playlist,” followed by entering any prompt in English. This feature is crafted to understand themes such as moods, aesthetics, and even experiences. Prompts can range from very general to highly specific, mentioning musical periods, genres, activities, lyrics, instruments, or even requesting a playlist themed around a TV series, film, or personal achievements. Users can also indicate whether they prefer the playlist to contain mainly new tracks or only selections from their library in the prompt.

After a prompt is put forward, Spotify’s AI curates a personalized playlist that fits the request. The system relies on the user’s listening history and blends it with contemporary music and cultural trends. In addition, each track includes a brief description explaining why it was chosen for that specific playlist.

Users can enhance their playlists by modifying their prompts or starting anew. For individuals whose musical preferences are ever-changing, playlists can be set to refresh automatically on a daily or weekly schedule.

As this feature is still in beta, Spotify indicated that adjustments may occur as the company processes feedback, and that currently there are usage restrictions. Some users have mentioned reaching limits after approximately 20 or 30 prompts.

Spotify has recently broadened its AI functionalities across its platform, including Page Match, which enables users to scan a physical book page to access the respective section in the audiobook, and About the Song. The platform also revised its song lyrics feature to include worldwide translations and offline availability. Last week, SeatGeek collaborated with Spotify to facilitate listeners in finding ticket links for concerts on an artist’s page or upcoming tour dates within the app. 

Internally, the company has integrated AI into its operations, with co-CEO Gustav Söderström stating earlier this month that Spotify’s top developers haven’t needed to write any code since December, thanks to AI.

Spotify is also enhancing its audiobook segment by moving into sales of physical books. Soon, users in the U.S. and U.K. will have the option to purchase physical editions directly through the app.