dera archive
Search the recent archive fast.
The archive is now focused on search and recent coverage. Timeline chapters and monthly reports live in Insights.
Recent coverage
Recent archive articles
Showing the latest available coverage.
Beyond Lovable and Mistral: 21 European startups to watch
Europe's startup scene is buzzing with more than just the usual suspects. TechCrunch, after consulting top VCs, has spotlighted 21 emerging companies, from pre-launch ventures to potential unicorns. Many of these leverage deep tech, particularly in AI, to solve specific industry problems, signaling Europe's unique approach to the global AI race.
Meta Introduces Autodata: An Agentic Framework That Turns AI Models into Autonomous Data Scientists for High-Quality Training Data Creation
Meta AI has introduced 'Autodata,' an innovative framework that transforms AI models into autonomous data scientists. This system allows AI to iteratively create and improve its own training and evaluation data, tackling the persistent challenge of data quality in AI development. By automating this crucial step, Autodata aims to accelerate the creation of more powerful and accurate AI models, potentially reducing development costs and time for businesses.
Musk v. Altman week 1: Elon Musk says he was duped, warns AI could kill us all, and admits that xAI distills OpenAI’s models
Elon Musk has initiated a lawsuit against OpenAI, claiming the company deviated from its founding non-profit principles. He alleges co-founders Sam Altman and Greg Brockman deceived him, transforming the organization into a for-profit entity. The legal battle also brings Musk's long-standing AI safety warnings to the forefront, even as details about his own AI venture, xAI, and its use of OpenAI's technology emerge in court.
OpenAI Enables Marketing Cookies by Default for Free ChatGPT Users
OpenAI recently changed its privacy policy, automatically enabling cookie tracking for free ChatGPT users. This move aims to help OpenAI promote its products and convert free users into paying subscribers. SMB leaders using the free version should review their privacy settings and consider opting out if they prefer not to be tracked for marketing purposes.
A Dark-Money Campaign Is Paying Influencers to Frame Chinese AI as a Threat
A US-based non-profit, 'Build American AI,' is running a large-scale social media campaign, framing China's AI progress as a national threat. This initiative, linked to a Super PAC supported by major tech leaders from OpenAI and Palantir, uses influencers to spread messages about potential risks to security and jobs, highlighting a growing rivalry in the AI space.
Pentagon inks deals with Nvidia, Microsoft, and AWS to deploy AI on classified networks
The U.S. Department of Defense recently announced new contracts with Nvidia, Microsoft, Amazon Web Services, and Reflection AI. These deals will allow their AI technologies and models to be deployed on classified networks for 'lawful operational use,' accelerating the military's transformation into an AI-first fighting force. This move signals the Pentagon's strategy to diversify its AI partners, aiming to enhance decision-making capabilities across all domains and avoid vendor lock-in.
Pentagon strikes classified AI deals with OpenAI, Google, and Nvidia — but not Anthropic
The U.S. Department of Defense has announced classified agreements with seven leading AI firms—OpenAI, Google, Microsoft, Amazon, Nvidia, xAI, and Reflection—to utilize their AI systems for national security. This move aims to establish the U.S. military as an AI-first fighting force. Notably, Anthropic, a previous contractor, was excluded due to supply chain risk concerns, despite its strong security model.
Anthropic Launches New Security Tool for Enterprises
Anthropic has launched 'Claude Security,' an AI-driven tool designed to scan enterprise codebases for vulnerabilities and generate fix patches. Built on the Claude Opus 4.7 model, this tool aims to bolster security defenses, especially for SMBs with limited dedicated security staff, by integrating seamlessly into existing workflows without requiring complex API setups.
Qwen AI Releases Qwen-Scope: An Open-Source Sparse AutoEncoders (SAE) Suite That Turns LLM Internal Features into Practical Development Tools
Ever wonder why your AI model sometimes goes off script, repeats itself, or rejects safe requests? Qwen AI has launched Qwen-Scope, an open-source suite designed to demystify Large Language Models (LLMs). By translating complex internal states into understandable concepts, Qwen-Scope empowers developers and SMBs to diagnose and steer AI behavior, leading to more reliable and controllable AI applications.
How Shivon Zilis Operated as Elon Musk’s OpenAI Insider
Recent court filings in the lawsuit between Elon Musk and OpenAI are pulling back the curtain on a fascinating early connection: Shivon Zilis. A long-time aide to Musk, Zilis served as an advisor and board member for OpenAI while also maintaining close ties to Musk. This dual role, now under legal scrutiny, reveals a critical communication channel that shaped OpenAI's early trajectory and Musk's influence.
Sources: Anthropic potential $900B+ valuation round could happen within 2 weeks
AI powerhouse Anthropic is reportedly pushing for an ambitious valuation of over $600 billion in its latest funding round. The company is asking investors to submit their allocations within 48 hours for a round expected to close within two weeks. This move highlights intense investor demand and sets a new benchmark in the competitive AI landscape.
Elon Musk's 7 biggest stumbles on the stand at OpenAI trial
Elon Musk's lawsuit against OpenAI hit a rough patch during his three-day testimony. Reports indicate that OpenAI's lawyers successfully challenged many of Musk's assertions, leading to several concessions and raising questions about his credibility on the stand.
After dissing Anthropic for limiting Mythos, OpenAI restricts access to Cyber, too
OpenAI has announced that its new cybersecurity testing tool, 'GPT-5.5 Cyber,' will initially be restricted to 'critical cyber defenders.' This decision is particularly notable because OpenAI CEO Sam Altman previously criticized competitor Anthropic for imposing similar access limitations on its own cybersecurity tool, 'Mythos,' calling it 'fear-based marketing.' The irony of OpenAI adopting a similar strategy is now a hot topic in the industry.
Mistral’s Model Lets You Vibe Long-Running Code in the Cloud
Mistral AI has launched its new 'Mistral Medium 3.5' model, integrated into a remote coding agent called 'Vibe' and an AI chatbot 'Le Chat.' These tools now feature a 'Work mode' that allows for extended, autonomous coding and various business tasks using natural language, promising significant improvements in development and operational efficiency for SMBs.
OpenAI announces new advanced security for ChatGPT accounts, including a partnership with Yubico
OpenAI just rolled out 'Advanced Account Security' (AAS) for ChatGPT users, a new optional feature designed to seriously beef up protection, especially for those handling sensitive info. Part of this big push is a collaboration with Yubico, bringing physical security keys like the YubiKey C NFC and YubiKey C Nano directly into the mix to shield ChatGPT accounts from growing phishing threats.
Elon Musk confirms xAI used OpenAI’s models to train Grok
Elon Musk's AI company, xAI, admitted to using OpenAI's models to train its Grok AI. This technique, called 'model distillation,' is widespread in the AI industry but operates in a legal gray area. While some see it as a legitimate way to create smaller, cheaper models, others view it as a shortcut that infringes on intellectual property, sparking debate among AI labs.
Elon Musk testifies that xAI trained Grok on OpenAI models
During a lawsuit against OpenAI, Elon Musk testified that his AI company, xAI, used a 'distillation' technique with OpenAI's models to train Grok. This admission brings to light a controversial practice of mimicking advanced AI models, previously thought to be more common among certain international competitors. The revelation challenges the competitive landscape among leading AI developers.
Google’s Gemini AI assistant is hitting the road in millions of vehicles
Google's Gemini AI assistant is coming to vehicles with integrated Google systems, marking a significant upgrade from the previous Google Assistant. This move promises more natural conversational AI for drivers, enabling enhanced vehicle control and information retrieval. The rollout begins in the US, with plans for broader availability and language support, and will reach both new and existing compatible vehicles via software updates.
Elon Musk Seemingly Admits xAI Has Used OpenAI's Models to Train Its Own
In a recent legal deposition, Elon Musk indicated that his AI company, xAI, may have used OpenAI's models to train its own AI. He explained 'distillation' as a common industry practice where one AI model helps train another, suggesting xAI's actions are standard. This revelation comes amid escalating competition and legal battles in the AI space, highlighting the complex dynamics of AI development and model utilization.
OpenAI Rolls Out ‘Advanced’ Security Mode for At-Risk Accounts
OpenAI has rolled out 'Advanced Account Security' for ChatGPT and Codex accounts. This new feature offers robust protection against phishing scams and account takeovers, especially critical for users handling sensitive information. It's a strategic move to secure AI services as they become more central to personal and business operations, mirroring advanced protection offered by tech giants like Google.
GitHub Copilot CLI for Beginners: Interactive v. non-interactive mode
GitHub has released a guide explaining the two main modes of Copilot CLI: interactive and non-interactive. This tool aims to integrate AI directly into the command line, promising to significantly enhance developer productivity. The article showcases practical applications, like generating emoji lists, and discusses how AI is reshaping software development, alongside the growing importance of understanding AI security.
Here’s how the new Microsoft and OpenAI deal breaks down
The complex relationship between Microsoft and OpenAI has taken a new turn, transitioning from an exclusive partnership to a more open arrangement. This 'friendly divorce' means OpenAI's models are now accessible on other cloud platforms like AWS, giving SMBs more flexibility in their AI adoption strategies and potentially intensifying competition in the AI market.
Gemini is rolling out to cars with Google built-in
Google announced it's bringing its powerful Gemini AI assistant to cars equipped with 'Google built-in.' This upgrade moves beyond simple commands, enabling more natural conversations, proactive suggestions, and access to vehicle-specific data like battery levels. It's set to roll out to both new and existing compatible vehicles, starting in the US, making in-car interactions much smarter and more intuitive for drivers.
This startup’s new mechanistic interpretability tool lets you debug LLMs
San Francisco startup Goodfire has launched 'Silico,' a new tool designed to peer inside AI models and adjust the parameters that govern their behavior. This innovation could help solve the 'black box' problem of LLMs, potentially improving AI development accuracy for SMBs who customize open-source models.
X announces a rebuilt ad platform powered by AI
X has officially launched its revamped advertising platform, now supercharged with AI. This move is a strategic play to re-engage advertisers and boost revenue after a period of decline. For SMBs, this could mean access to more precise targeting and automated campaign management, potentially leveling the playing field with larger competitors.
OpenAI talks about not talking about goblins
OpenAI's AI models began unexpectedly referencing 'goblins' and 'gremlins,' sparking industry discussion. The company now explains this 'goblin problem' stemmed from a 'Nerdy' personality option that inadvertently rewarded such metaphors, influencing subsequent model training. While the personality is gone, remnants persist, especially in Codex, requiring direct intervention.
Meta says its business AI now facilitates 10 million conversations a week
Meta's business-focused AI tools are experiencing explosive growth, hitting 10 million weekly conversations, up from 1 million earlier this year. This surge is fueled by Meta's strategic move to offer these AI assistants for free to SMBs across various regions. The tools, especially for ad creative, are proving effective, with some advertisers seeing higher conversion rates, signaling a significant shift in how businesses can leverage AI for marketing.
OpenAI’s new security model is for ‘critical cyber defenders’ only
OpenAI is launching GPT-5.5-Cyber, a new advanced cybersecurity model, exclusively for 'cyber defenders' rather than the general public. This move aligns with an industry trend where top-tier AI models are being released incrementally due to potential misuse risks. While details on its capabilities are scarce, it's expected to be a specialized version of the powerful GPT-5.5, aimed at strengthening organizational defenses.
SoftBank is creating a robotics company that builds data centers — and already eyeing a $100B IPO
SoftBank is reportedly creating a new company, Roze AI, to automate data center construction, especially in the US. This venture aims to use autonomous robots to build server farms more efficiently. Interestingly, SoftBank is already preparing for a potential $100 billion IPO for Roze AI by late 2026, though some internal skepticism exists given past high-profile failures in AI investments.
Sources: Anthropic could raise a new $50B round at a valuation of $900B
Anthropic, the company behind the AI assistant Claude, is reportedly eyeing a massive new funding round of approximately $50 billion at a staggering $900 billion valuation. This move, coming hot on the heels of a $380 billion funding round in February, highlights the intense and rapidly escalating investment frenzy in the AI market, placing Anthropic's valuation on par with or even exceeding that of rivals like OpenAI.