Latest AI News
Guide Labs debuts a new kind of interpretable LLM
The challenge of wrangling a deep learning model is often understanding why it does what it does: Whether it’s xAI’s repeated struggle sessions to fine-tune Grok’s odd politics, ChatGPT’s struggles with sycophancy, or run-of-the-mill hallucinations, plumbing through a neural network with billions of parameters isn’t easy. Guide Labs, a San Francisco start-up founded by CEO Julius Adebayo and chief science officer Aya Abdelsalam Ismail, is offering an answer to that problem today. On Monday, the company open-sourced an 8 billion parameter LLM,Steerling-8B, trained with a new architecture designed to make its actions easily interpretable: Every token produced by the model can be traced back to its origins in the LLM’s training data. That can as a simple as determining the reference materials for facts cited by the model, or as complex as understanding the model’s understanding of humor or gender. “If I have a trillion ways to encode gender, and I encode it in 1 billion of the 1 trillion things that I have, you have to make sure you find all those 1 billion things that I’ve encoded, and then you have to be able to reliably turn that on, turn them off,” Adebayo told TechCrunch. “You can do it with current models, but it’s very fragile … It’s sort of one of the holy grail questions.” Adebayo began this work while earning his PhD at MIT, co-authoring a widely cited2018 paperthat showed existing methods of understanding deep learning models were not reliable. That work ultimately led to the creation of a new way of building LLMs: Developers insert a concept layer in the model that buckets data into traceable categories. This requires more up front data annotation, but by using other AI models to help, they were able to train this model as their largest proof of concept yet. “The kind of interpretability people do is…neuroscience on a model, and we flip that,” Adebayo said. “What we do is actually engineer the model from the ground up so that you don’t need to do neuroscience.” One concern with this approach is that it might eliminate some of the emergent behaviors that make LLMs so intriguing: Their ability to generalize in new ways about things they haven’t been trained on yet. Adebayo says that still happens in his company’s model: His team tracks what they call “discovered concepts” that the model discovered on its own, like quantum computing. Adebayo argues this interpretable architecture will be something everyone needs. For consumer-facing LLMs, these techniques should allow model builders to do things like block the use of copyrighted materials, or better control outputs around subjects like violence or drug abuse. Regulated industries will require more controllable LLMs, for example in finance, where a model evaluating loan applicants needs to consider things like financial records but not race. There’s also a need for interpretability in scientific work, another area where Guide Labs has developed technology. Protein folding has been a big success for deep learning models, but scientists need more insight into why their software figured out promising combinations. “This model demonstrates is that training interpretable models is no longer a sort of science; it’s now an engineering problem,” Adebayo said. “We figured out the science and we can scale them, and there is no reason why this kind of model wouldn’t match the performance of the frontier level models,” which have many more parameters. Guide Labs says that Steerling-8B can achieved 90% of the capability of existing models, but uses less training data, thanks to its novel architecture. The next step for the company, which emerged from Y Combinator and raised a $9 million seed round from Initialized Capital in November 2024, is to build a larger model and begin offering API and agentic access to users. “The way we’re current training models is super primitive, and so democratizing inherent interpretability is actually going to be a long term good thing for our role within the human race,” Adebayo told TechCrunch. “As we’re going after these models that are going to be super intelligent, you don’t want something to be making decisions on your behalf that’s sort of mysterious to you.”
View
OpenAI calls in the consultants for its enterprise push
OpenAI is beefing up partnerships with four major consulting giants as the AI company looks to grow its enterprise business in 2026. OpenAI announced on Monday the“Frontier Alliance,”a signal that the AI lab is willing to try different approaches to get enterprises to meaningfully adopt its technology. The alliance includes multi-year partnerships between OpenAI and four major consulting firms, Boston Consulting Group (BCG), McKinsey, Accenture and Capgemini, to sell its enterprise products. OpenAI’s Forward Deployed Engineering team will work with the consulting giants to help them implement OpenAI’s enterprise-focused technologies like OpenAI Frontier into customers’ tech stacks. The companylaunched OpenAI Frontier in early February. The no-code open software allows users to build, deploy, and manage AI agents both built on OpenAI’s AI models and beyond. OpenAI argues in its latest announcement that consultants are the right avenue to get enterprises on board. “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 said in OpenAI’s blog post. “Our expanded partnership combines OpenAI’s Frontier platform with BCG’s deep industry, functional, and tech expertise and BCG X’s build-and-scale capabilities to drive measurable impact with safeguards from day one.” Thus far,enterprise adoption of AI has been relatively slowas these companies struggle to find a meaningful return on investment from their AI pursuits. OpenAI’s alliance strategy makes sense and goes beyond just pitching enterprises on attaching AI to their existing workflows. Instead, this effort focuses on consultants persuading companies to change their strategies and workflows to fold in OpenAI’s tools where it makes sense. It’s worth noting the OpenAI rival Anthropic has inked deals with consulting giants includingDeloitteandAccenturein recent months too. Company CFO Sarah Friar wrote in a blog post in January that enterprise is abig area of focus for OpenAIin 2026. OpenAI has also inked sizable enterprise AI deals withSnowflakeandServiceNowso far this year, in addition to naming Barret Zoph to lead the company’s enterprise sales effort in January.
View
Samsung Galaxy S26 Series to Support 'Hey Plex' Hotword for Perplexity Voice Assistant
Samsung on Monday announced that the upcoming Galaxy S26 series of smartphones, which is expected to comprise the standard Galaxy S26, Galaxy S26+, and the Galaxy S26 Ultra, will offer support for another AI agent. An addition to Galaxy AI, the Samsung Galaxy S26 series will also support Perplexity AI, according to the company. Users will be able to invoke the AI assistant using the 'Hey Plex' hotword, or by pressing the side button. It will also work across the company's built-in apps as well as some third-party apps.
View
Honor Teases Humanoid Robot, Schedules Showcase a Day Ahead of MWC 2026
The Mobile World Congress (MWC 2026) is scheduled to begin on March 2 and will go on until March 5. During the four-day event, various smartphone makers and tech firms will showcase their new products and innovations. Ahead of MWC 2026, various Chinese companies, including Xiaomi and Honor, will also globally unveil new devices. Recently, Honor confirmed that its Magic V6 foldable and the Robot Phone with a pop-out camera will be globally launched a day before the event starts. Now, the Chinese smartphone maker has confirmed that the handsets will be accompanied by a humanoid robot.
View
Defense Secretary summons Anthropic’s Amodei over military use of Claude
Defense Secretary Pete Hegseth is calling in Anthropic CEO Dario Amodei to the Pentagon on Tuesday morning to discuss the military use of Claude, according to reporting fromAxios. The meeting comes as the Pentagon threatens to declare Anthropic a “supply chain risk” — a label typically reserved for foreign adversaries — after the AI firm refused to allow the Department of Defense to use its tech for the mass surveillance of Americans and the development of weapons that fire without human involvement. Anthropic signed a $200 million contract with DOD last summer, and Claude wasreportedlyused during the January 3 special operations raid that resulted in the capture of Venezuelan president Nicolás Maduro, an episode that brought the two sides’ tensions into the open. A source told Axios that Hegseth is giving Amodei an ultimatum: play ball or be banished. It’s unclear whether he’s bluffing — replacing Anthropic would be a significant undertaking. But the stakes are real: a supply chain risk designation would void Anthropic’s contract and force other Pentagon partners to drop Claude entirely.
View
How AI agents could destroy the economy
On Sunday, an analyst group called Citrini Research publisheda remarkable pieceillustrating how agentic AI could bring on mass economic destruction over the next two years. The scenario imagines a report from two years in the future, in which unemployment has doubled, and the total value of the stock market has fallen by more than a third. As the report puts it: AI capabilities improved, companies needed fewer workers, white collar layoffs increased, displaced workers spent less, margin pressure pushed firms to invest more in AI, AI capabilities improved… It was a negative feedback loop with no natural brake…The system turned out to be one long daisy chain of correlated bets on white-collar productivity growth. It’s a new kind of bear case, focused not on Skynet-style misalignment but on the gradual unspooling of the economy itself. In particular, the Citrini scenario looks at the implications of integrating AI agents into the economy at large, and what it would mean when outside contractors get replaced by cheaper in-house AI. It’s similar to theDeath of SaaSscenario, but Citrini goes further, implicating any business model that involves optimizing transactions between companies. As you might expect, the report is causingquiteastironline. Not everyone is buying it — even Citrini describes it as more of a scenario than a prediction — but it’s not so easy to name the specific point where you think the scenario goes wrong. Personally, I’m not sure companies are ready to hand off purchasing decisions to AI agents, no matter how smart they are. But in Citrini’s scenario, most of the impacted decisions have already been handed off to third-party contractors, so it’s not quite as implausible as it seems.
View
Top 15 AI Startups Powering India’s Self-Reliance Mission
From foundational models to applied platforms, domestic builders are creating an end-to-end AI ecosystem rooted in Indian data and needs.
View
Amazon Opens Second-Largest Asia Office in Bengaluru
The 1.1 million square feet campus, spread across 12 floors on a five-acre site, can accommodate more than 7,000 employees.
View
LTM Lands $100 Mn Deal With European Medtech to Support Hearing Devices
LTM’s pact with a European medtech marks its first win after the rebrand.
View
AI May Reset Indian IT’s Pyramid Model, But Not Fresher Hiring
Industry voices repeatedly stress skills evolution, AI literacy and the accelerating pace of learning.
View
‘We are Not in the Race to Build Data Centres’: Nikhil Malhotra
The Tech Mahindra CIO spoke to AIM about Project Indus and why the company decided to outsource heavier compute.
View
This 23,000-Employee MNC is Betting on Pune Nano GCC for AI Solutions
Orbia’s Pune GCC houses roughly one-third of its global IT talent pool.
View