AI News Archive: August 6, 2026 — Part 1
Sourced from 500+ daily AI sources, scored by relevance.
- Scientists Trained An AI Model In DNA—And It Invented 16 New Viruses
Scientists are worried about how smart DNA-trained artificial intelligence models could get—and that they could become a weapon in biological warfare.
- OpenAI Didn’t Notice Its AI Agents Using a Message Board to Plan Their Hacking Spree
At the Black Hat security conference, the AI giant revealed new details about how its agents went rogue, hacked several other companies—and did it all right under the company’s nose.
- Google shifts AI power back to Brin as DeepMind’s Hassabis steps aside
Silicon Valley parent consolidates control as the London lab’s scientific culture gives way to urgency to build AI products
Score: 92🌐 MovesAug 6, 2026https://www.ft.com/content/1453e9c2-4922-482f-8720-0bafd7e07df7?syn-25a6b1a6=1 - Improving GPT‑5.6 Sol in ChatGPT—and expanding access to GPT-5.6 Luna for free users
ChatGPT introduces improved GPT-5.6 Sol with better accuracy and consistency, plus expanded access for free users and unlimited everyday chats with GPT-5.6 Luna.
- Exclusive: Stripe in Exclusive Talks to Buy Startup OpenRouter for Around $10 Billion
Exclusive: Stripe in Exclusive Talks to Buy Startup OpenRouter for Around $10 Billion The Information
- Software Giant SAP Stops Most Travel and Hiring Because of AI’s Soaring Cost
SAP says it needs to “be disciplined in how we spend.” That includes still freezing hires and travel. Unless it's to do with AI, of course.
Score: 89🌐 MovesAug 6, 2026https://www.404media.co/software-giant-sap-stops-most-travel-and-hiring-because-of-ais-soaring-cost/ - No cloud, no GPUs, no problem: Liquid AI's new model LFM2.5-2.6B brings powerful AI agents to devices as small as a Raspberry Pi
Earlier this week, the AI startup Liquid, formed in 2023 by former MIT computer scientists, debuted LFM2.5-2.6B , a new open-weight language model designed specifically for agentic workloads. In release materials and a recent interview with VentureBeat, Liquid's researchers said LFM2.5-2.6B can run entirely on local hardware — from smartphones and laptops down to a Raspberry Pi — without relying on cloud inference or GPUs, unlocking edge AI applications and giving more options to enterprises working in regulated industries or with sensitive information they don't want to send up to the cloud. It's best suited for high-volume, well-defined agentic tasks that run locally — tool calling, document management, calendar and workflow automation, and always-on background routines — and for connectivity-limited environments like vehicles and robotics, though coding-heavy work is better left to larger models. Even for those businesses without such concerns, the appeal of running performant, task-specific agents at the cost of essentially electricity, may be enough to make the new model quite appealing. But the custom open weights license , as with Moonshot's larger frontier model Kimi K3 released last month, is worth a close look by enterprise legal teams. The basics LFM2.5-2.6B contains 2.6 billion parameters, supports a 128,000-token context window, and includes native tool calling. The somewhat tricky name is explained by the generation of model (2.5) combined with the parameter count (2.6B). Both the post-trained model and a base checkpoint (LFM2.5-2.6B-Base) for developers who want to fine-tune it are available now on Hugging Face , with day-one support for major inference stacks including llama.cpp, MLX, vLLM, SGLang, and ONNX — positioning it for deployment across consumer hardware, enterprise infrastructure, and embedded systems. Liquid also offers an open source fine-tuning framework, LEAP . Rather than positioning LFM2.5-2.6B as a competitor to the largest frontier models, the company is making a different argument: that a sufficiently capable small model can unlock categories of enterprise applications where latency, privacy, deployment flexibility, or inference costs matter more than absolute benchmark leadership. "I do also believe that the best models will be in the cloud, and there's no problem with that," Maxime Labonne, Liquid AI's head of post-training, told VentureBeat in an interview following the launch. "We want to make models for another type of user, and the best way of describing it is: you should use [edge AI] when you can't use a cloud model." Small enough for a Raspberry Pi Asked about the minimum viable hardware, Labonne said the model runs "very, very well" on CPUs — and that the LFM2 architecture underlying the model was explicitly designed around real-world CPU performance rather than GPU benchmarks. "I think the best example is a Raspberry Pi," he said. "We have a lot of demos that show that actually, it works pretty fast on the Raspberry Pi." Company-reported measurements indicate decoding throughput of approximately 220 tokens per second on an Apple M5 Max and 113 tokens per second on an AMD Ryzen AI Max+ 395, while using less than 2.5 GB of memory — and around 30 tokens per second on a smartphone. Users can try the models on their phones through Apollo, Liquid AI's mobile app. At the other end of the deployment spectrum, Liquid AI reports the model reaches nearly 15,000 output tokens per second on a single Nvidia H100 GPU under sustained concurrent load — roughly 1.3 billion tokens per day on one card. These figures are vendor benchmarks and have not been independently verified. For Labonne, memory footprint and speed are not conveniences but hard constraints that determine what can be deployed at all. "What we want to show is that it's a really good trade-off, because you get the level of quality that you get with much bigger models, but in a tiny, tiny form factor," he said. "You can deploy it in target devices where you are not able to deploy the other ones at all." Trained for agents instead of chatbots Liquid AI says LFM2.5-2.6B was developed around the assumption that language models are increasingly consumed through agent frameworks rather than traditional conversational interfaces. "Models are not consumed in chatbots anymore. They're really consumed through agentic harnesses, like OpenClaw, like Hermes Agent," Labonne said. "We wanted to make sure that this model is not just good at math or at code, but it's good at using tools." The model is pretrained on approximately 34 trillion tokens, with a vocabulary doubled to 128K to better support non-Latin scripts and a dedicated mid-training phase to extend the context window to 128K tokens for long-running agent workflows. Post-training follows a four-stage pipeline: supervised fine-tuning, teacher specialization (training separate expert models for domains like instruction following, math, code, and tool use), multi-domain on-policy distillation (MOPD) to merge those experts' capabilities back into a single student model, and finally agentic reinforcement learning. During that last stage, the model was trained directly inside production agent harnesses — including Hermes Agent and OpenClaw — on realistic productivity tasks involving research, coding, document management, tool invocation, and workflow automation, exposing it to those harnesses' actual tools, system prompts, and interaction patterns. Labonne described the pipeline overhaul as producing a "happy accident": gains that extended well beyond the agentic targets. "Through these new training techniques, we also got a lot better at everything. We got better at math, at instruction following. We've never been good at code, actually — and with this, we even got really good at code," he said. Building the model — and the harness Notably, Liquid AI also built its own agent harness rather than relying solely on existing frameworks, and demonstrated the model running inside it on a phone, planning and calling tools entirely on-device. "This is a harness running on a phone, and I don't know if there's any other harness running on a phone," Labonne said. The company had two reasons, he explained. The first was necessity — no phone-native harness existed. The second is a different interaction model: today's harnesses wait for a prompt, and Liquid AI wants assistants that act on their own. "We want proactive agents. We want agents that run in the background, check what you're doing, check your calendar, and based on this context, do tasks," he said. "That doesn't exist today, really." Co-designing the harness and model also lets the software compensate for the model's weak spots. "Everything that the model is bad at, the harness should help the model with — provide as much assistance as possible to make it more reliable," Labonne said. "End users don't care if it's the model or the harness. What they want is that the task is achieved at the end of the day." The model nevertheless works out of the box with established harnesses including Hermes Agent, OpenClaw, and Pi, served behind any OpenAI-compatible endpoint. Swap the harness, not the model For enterprise deployment, Labonne argued the release marks a shift in what small models can be used for. Until now, he said, local models made economic sense mainly as narrowly fine-tuned specialists — trained to do one thing at cloud-model quality, much faster and cheaper. Agentic capability changes that calculus, because the same model can be repurposed by changing the tools around it rather than the model itself. "You can have a calendar assistant, and you can reuse the same model and make a meeting assistant that will record what everybody said and summarize it — a bit like Granola, for example," he said. "You don't change the model; you just change the harness. You just change the tools around it. This gives much more generalizability, and it's a lot easier to do and a lot cheaper as well." He still recommends fine-tuning for production deployments whenever feasible: "If you don't fine-tune it, you leave some quality on the table. If you fine-tune it well, it's going to match the performance of GPT and Claude — really, if your task is not the most complex task in the world," he said, adding that the barrier to entry has collapsed: "The bar to be able to do fine-tuning now is super low. It's very accessible to everyone." How it stacks up against DeepSeek-V4-Flash, Google's Gemma and Alibaba's Qwen Liquid AI released its own benchmark comparison charts pitting LFM2.5-2.6B against the models enterprises are most likely to shortlist for the same edge deployments: Google's Gemma 4 E2B (5.1B parameters) and E4B (8B), and Alibaba's Qwen3.5-4B (4.7B) and Qwen3.5-9B (9.7B). A separate test by local AI client platform Atomic Chat found that LFM2.5-2.6B completed 35 tool calls to complete three tasks (checking weather and local time in six cities, converting one budget into six currencies, checking four hotels and booking for a date) 3.7 times faster than DeepSeek-V4-Flash (a whopping 284B parameters), the model has skyrocketed to the top of OpenRouter since its release last week. Gemma 4's small models are multimodal generalists, accepting image and audio input alongside text, and use a Per-Layer Embeddings design that keeps only a fraction of their weights active per token — which is why Google markets them by "effective" size (2.3B and 4.5B) despite total footprints of 5.1B and 8B. Alibaba's Qwen3.5 small series, released in March , is natively multimodal from 4B up and leans on scaled reinforcement learning to chase frontier-style reasoning — Alibaba touts the 9B model as matching or beating OpenAI's far larger gpt-oss-120B on reasoning benchmarks. LFM2.5-2.6B takes a narrower path: it is text-only, dense, and specialized for agentic work, with Liquid AI shipping separate vision and audio variants of the LFM family rather than folding everything into one checkpoint. Where Qwen's post-training reinforcement learning targets reasoning, Liquid's targets tool use inside real agent harnesses. The result, per the company's published numbers, is that the smallest model in the comparison leads every instruction-following benchmark (IFBench, Multi-IF, IFStruct) and nearly every tool-use benchmark — 77.83 on ToolSandbox versus 76.44 for Qwen3.5-9B, a model nearly four times its size — trailing only that 9B model on BFCLv4. On agentic evaluations it beats both Gemma models across the board and essentially ties the Qwens: 26.89 on BrowseComp+ versus 27.23 for Qwen3.5-9B. It also posts the best score on AA Omniscience, a knowledge benchmark that penalizes hallucination. The Qwen models keep the edge where their training focus lies: math (Qwen3.5-9B leads AIME25) and coding, where larger models retain an advantage on LiveCodeBench — though Labonne noted the gap is smaller than the parameter counts would suggest. "With LiveCodeBench v6, we might not be the best among these models, but we're also by far the smallest. Showing that we're competitive with them is already quite a big win for me," he said. One differentiator cuts the other way: licensing. Gemma 4 and Qwen3.5 ship under the permissive Apache 2.0 license — a change Google made specifically to court enterprises . DeepSeek-V4-Flash ships under a similarly permissive MIT License . Meanwhile, Liquid AI's revenue-gated license (detailed below) asks larger companies to strike a commercial deal. Enterprises above the threshold are effectively trading license friction for footprint and tool-use performance. Licensing reflects a commercial middle ground LFM2.5-2.6B is distributed under the LFM Open License v1.0, which permits use, modification, and redistribution — including commercial use — for organizations with less than $10 million in annual revenue. Commercial use by larger companies is not covered by the license, requiring a separate arrangement with Liquid AI; qualified nonprofits are exempt from the threshold for non-commercial and research purposes. Labonne framed the structure as a way to sustain model development — "the models are really the moats, so we need to be sensible in the way that we license them; otherwise, we cannot make money, so we can't make more models" — while characterizing the threshold as a light-touch mechanism in practice. Asked how the company would even know if a large enterprise quietly deployed the open weights, he was candid: "I think this is a question for our legal team, but personally, I don't know. And even if you're above $10 million, the only thing that we ask you is to contact us." The company pairs its licensed model releases with freely published research, he added, including new structured-output evaluations and a training technique that mitigates the repetition loops common in small models — a failure mode he noted Qwen models are "kind of guilty of." Small model, big enterprise implications The launch coincided with an announcement from MacPaw , the Ukrainian software company behind CleanMyMac and Setapp, of a long-term strategic partnership with Liquid AI to build an on-device AI stack for the Mac. Liquid AI will design and fine-tune foundation models for Eney, MacPaw's macOS assistant, running locally on Apple silicon through MacPaw's Elix inference engine and Mnemos memory layer, with results expected later this year. Labonne pointed to the deal as a concrete validation of the size argument: "One of the reasons why they chose us is also because the model is quite small, and they don't have all the memory budget to run the other models." The release arrives as hardware vendors, operating system developers, and enterprise software companies increasingly invest in local AI execution — and as agent harnesses proliferate across the industry. Liquid AI's bet is that deployment economics, not raw scale, will define an important segment of that market: agents running continuously, everywhere, at zero marginal token cost. Whether small, highly optimized agent models become a significant segment of enterprise AI will ultimately depend less on benchmark scores than on operational reliability. But Liquid AI's latest release suggests the next competitive frontier is no longer simply building larger models — it's building models small enough, and capable enough, to run wherever enterprise workflows already live.
- Groq Accelerates COVID Drug Discovery by 333x for Argonne National Lab
Groq Accelerates COVID Drug Discovery by 333x for Argonne National Lab
Score: 87🌐 MovesAug 6, 2026https://groq.com/blog/groq-accelerates-covid-drug-discovery-333x-versus-legacy-solutions - Wayve & Uber Can Now Deploy Robotaxis in London
The year of robotaxis moves forward, and we have another robotaxi update today. The autonomous driving firm Wayve and partner Uber were granted Private Hire Vehicle licences from Transport for London (TfL) and can now deploy robotaxis in The Swinging City. That said, they must include human supervisors onboard for ... [continued] The post Wayve & Uber Can Now Deploy Robotaxis in London appeared first on CleanTechnica .
Score: 86🌐 MovesAug 6, 2026https://cleantechnica.com/2026/08/05/wayve-uber-can-now-deploy-robotaxis-in-london/ - AI system designs functional bacteriophages from scratch
AI system designs functional bacteriophages from scratch eurekalert.org
- Google targets AI startup Mechanize’s technology and talent in proposed $1.5B deal
Google LLC is reportedly in discussions with an artificial intelligence coding startup called Mechanize Inc. over a deal that would see it license some of its technology and hire a number of its most talented employees. Business Insider was the first to report the discussions, citing information from four people familiar with the conversations. Mechanize was […] The post Google targets AI startup Mechanize’s technology and talent in proposed $1.5B deal appeared first on SiliconANGLE .
- Meta AI Model Hacked Outside Company, Adding to Concerns Over Rogue Bots
During cybersecurity testing the model slipped onto the internet and hacked a third-party service, mirroring incidents at Anthropic and OpenAI.
- AMD buys chip startup that hardwires AI models into its silicon
Taalas' current chip runs a small version of Meta's Llama 3.1, though the company is working on chips for bigger and more advanced models.
Score: 84💰 MoneyAug 6, 2026https://www.cnbc.com/2026/08/06/amd-buys-taalas-startup-that-hardwires-ai-models-into-its-silicon.html - Alphabet seeking up to $25-billion in fresh debt to finance AI build-out
Tech giants increasingly turning to bond markets to fund AI spending
Score: 84💰 MoneyAug 6, 2026https://www.theglobeandmail.com/investing/article-alphabet-25-billion-bond-sale-ai-build-out/ - Claude Fable 5 AI finds a tiny formula that topples an 87-year-old math conjecture
A mathematician working at Anthropic says he used the AI model Claude Fable 5 to uncover a remarkably simple counterexample to the Jacobian conjecture, a famous problem that has resisted mathematicians for more than a century. The result shows that the conjecture is false in three dimensions and above, although the original two-dimensional version remains unsolved.
- SpaceX, Tesla to initially spend $16.8 billion on Terafab chip plant in Texas
SpaceX, Tesla to initially spend $16.8 billion on Terafab chip plant in Texas Reuters
- GPT-5 turning one as OpenAI shares new Agent Plugins standard
OpenAI’s GPT-5 turns one tomorrow, and the company is marking the week by looking beyond individual models. Today, OpenAI introduced Agent Plugins, an open standard meant to let reusable AI-agent extensions work across compatible products.
Score: 83🤖 ModelsAug 6, 2026https://9to5mac.com/2026/08/06/gpt-5-turning-one-as-openai-shares-new-agent-plugins-standard/ - Qwen 3.8 Max’s Incredible Debut: Better and Cheaper
A model ran unsupervised for ten straight days, building and testing its own software harness. Continue reading on Towards AI »
- Groq Partners with Aramco on World’s Largest AI Data Center
Groq Partners with Aramco on World’s Largest AI Data Center
- SoftBank Uses OpenAI Stake to Borrow $10 Billion
The Japanese technology investment company has vowed to become a leading builder of AI data centers and a major investor in AI-powered robotics.
- DeepSeek V4 Flash 0731 Outscores V4 Pro at 5x Lower Cost
DeepSeek V4 Flash 0731 Beats Its Own Pro Preview at $0.14 a Million Tokens That’s the Real Story Continue reading on Towards AI »
- China’s Unitree Seeks $904 Million in First Mainland Robotic IPO
China’s Yushu Technology Co., known as Unitree Robotics, is seeking to raise about 6.1 billion yuan ($904 million) in an initial public offering that would make it the mainland’s first publicly traded humanoid robot maker.
- Alibaba’s latest AI model puts it back in the great game
The ecommerce and cloud company hopes to recoup some of its lost shine with Monday’s release
Score: 81🤖 ModelsAug 6, 2026https://www.ft.com/content/391c5f14-4bf2-4d24-92f0-65b912574513?syn-25a6b1a6=1 - Anthropic will design its own hardware to power Claude
Anthropic and OpenAI are racing to scale up while reducing dependence on Nvidia.
Score: 80🌐 MovesAug 6, 2026https://arstechnica.com/ai/2026/08/anthropic-confirms-plans-to-build-an-in-house-silicon-team/ - Acrab Raises US$130 Million Series B, Advancing Agentic AI Compute Platform Commercialization
Acrab Raises US$130 Million Series B, Advancing Agentic AI Compute Platform Commercialization The Straits Times
- UAE fund weighs $6.3 billion AI data center investment in Japan
UAE fund weighs $6.3 billion AI data center investment in Japan The Japan Times
- How OpenAI Is Capturing the Most Valuable Moment in Healthcare
OpenAI is capitalizing on the moment people seek health guidance—before they ever reach a clinic.
Score: 78🌐 MovesAug 6, 2026https://www.inc.com/chase-feiger/how-openai-is-capturing-most-valuable-moment-in-healthcare/91384537 - AMD deepens AI inference bet with Taalas deal as chip race heats up
AMD deepens AI inference bet with Taalas deal as chip race heats up Reuters
- Columbus AI welding company wins $600M contract with largest U.S. shipbuilder
Columbus-based Path Robotics and California’s GrayMatter Robotics will develop an autonomous production line for U.S. Navy contractor HII.
- Waymo opens fully autonomous robotaxi service to all Dallas residents after delay
Waymo's fully autonomous robotaxis are now available to all Dallas residents after months of weather and safety concerns delayed the public launch, making Dallas the second Texas city where anyone can hail a driverless ride.
Score: 78🌐 MovesAug 6, 2026https://www.bizjournals.com/dallas/news/2026/08/06/waymo-opens-rides-to-all-in-dallas.html?ana=brss_6150 - Indosat launches AI infrastructure venture with Ooredoo, NVIDIA and Nokia
Indosat launches AI infrastructure venture with Ooredoo, NVIDIA and Nokia DealStreetAsia
Score: 78🌐 MovesAug 6, 2026https://www.dealstreetasia.com/stories/indosat-ooredoo-nvidia-nokia-491445/ - DeepSeek invests $20.8 million in Unitree's Shanghai IPO
DeepSeek invests $20.8 million in Unitree's Shanghai IPO Reuters
Score: 77💰 MoneyAug 6, 2026https://www.reuters.com/world/asia-pacific/deepseek-invests-208-million-unitrees-shanghai-ipo-2026-08-06/ - Fare Is Fair: Zoox Launching Robotaxi Service in Las Vegas
Optimized fares will be based on the best route from pickup to drop-off, with the full price shown before you book and no surprises if the robotaxi takes a longer way around.
Score: 77🌐 MovesAug 6, 2026https://www.cnet.com/roadshow/automobiles/zoox-paid-robotaxi-service-las-vegas-aug-10/ - NVIDIA Leads USD 74 Million Investment in Sarvam AI
NVIDIA Leads USD 74 Million Investment in Sarvam AI india.entrepreneur.com
Score: 77💰 MoneyAug 6, 2026https://india.entrepreneur.com/business-news/nvidia-leads-usd-74-million-investment-in-sarvam-ai - Aurora launches second-generation driverless trucks across 10 Sun Belt freight routes
Aurora Innovation has deployed its second generation of fully driverless Class 8 trucks on commercial freight routes, putting autonomous hardware built for a one-million-mile service life into revenue-generating service across the U.S. Sun Belt. New platform, lower costs, broader routes Aurora Innovation has launched its second-generation driverless trucks, with plans to deploy the new fleet ... Read more
Score: 76🌐 MovesAug 6, 2026https://gcn.com/aurora-launches-second-generation-driverless-trucks/20501/ - USC pushes to expand AI research with $200 million gift
USC pushes to expand AI research with $200 million gift marketplace.org
Score: 76🌐 MovesAug 6, 2026https://www.marketplace.org/story/2026/08/06/usc-pushes-to-expand-ai-research-with-200-million-gift - Envision Commissions Galaxy Campus in Ulanqab, Establishing a New Model for Gigawatt-Scale AI Infrastructure
Envision Commissions Galaxy Campus in Ulanqab, Establishing a New Model for Gigawatt-Scale AI Infrastructure The Straits Times
- Google open-sources an AI model it says can help with earlier hurricane warnings
WeatherNext can deliver a 15-day forecast predicting storms' track and intensity.
Score: 76🤖 ModelsAug 6, 2026https://www.engadget.com/2232141/google-open-source-ai-model-can-help-with-earlier-hurricane-warnings/ - AI-empowered pipeline-in-a-drug option: Insilico Medicine nominates ISM9077, potential first-in-class target Y inhibitor, as preclinical candidate (PCC) for ocular diseases, inflammatory disorders and aging
AI-empowered pipeline-in-a-drug option: Insilico Medicine nominates ISM9077, potential first-in-class target Y inhibitor, as preclinical candidate (PCC) for ocular diseases, inflammatory disorders and aging eurekalert.org
- Miami startup lands $25M to bring AI into remote patient monitoring for chronic disease
The company's technology is already used by major health systems.
- PokeBot Raises Hundreds of Millions of Dollars in Pre-A Funding
PokeBot, a Chinese embodied-AI startup, has raised hundreds of millions of dollars in a pre-A round led by Shunwei Capital and Matrix Partners. The months-old company is developing household robots for tasks including clothes folding and cooking. A demonstration showed one of its robots autonomously cooking mapo tofu in nine minutes. [TechNode reporting]
Score: 75💰 MoneyAug 6, 2026https://technode.com/2026/08/06/pokebot-raises-hundreds-of-millions-of-dollars-in-pre-a-funding/ - Dronetech Startup Garuda Aerospace Gets SEBI Nod For ₹750 Cr+ IPO
Chennai-based Garuda Aerospace has received SEBI’s final observation for its proposed IPO, paving the way for the dronetech startup to…
Score: 75💰 MoneyAug 6, 2026https://inc42.com/buzz/dronetech-startup-garuda-aerospace-gets-sebi-nod-for-%e2%82%b9750-cr-ipo/ - A startup that places ads for brands like Best Buy and Target in AI platforms raised $30.5 million
A startup that places ads for brands like Best Buy and Target in AI platforms raised $30.5 million Business Insider
Score: 75💰 MoneyAug 6, 2026https://www.businessinsider.com/ai-chatbot-ad-platform-gravity-raises-series-a-2026-8 - Naïve raises $28.5M to automate the grunt work of setting up and running a company
Taking vibe-coding a step further, Naïve claims its infra can automate most of the work in setting up and running a business.
- Top Google minds quit to build AI that accelerates research
Discovery Loop wants to advance the pace of scientific discovery using AI to address issues across domains. Read more: Top Google minds quit to build AI that accelerates research
Score: 75🌐 MovesAug 6, 2026https://www.siliconrepublic.com/start-ups/top-google-minds-leave-discovery-loop-jeff-dean - Meta launches Muse Code for complex software work with persistent AI agents
Meta launches Muse Code for complex software work with persistent AI agents InfoWorld
- How OpenAI's agents broke out of testing to hack Hugging Face
Weeks before OpenAI's agents hacked Hugging Face , the agents worked together to find and exploit a vulnerability in the infrastructure supporting the company's cybersecurity testing, OpenAI researchers said Wednesday. Why it matters: The new findings raise questions about how frontier AI labs are monitoring their testing environments — and the challenges safety testers are finding as they try to rein in increasingly powerful AI. Driving the news: OpenAI's internal research model, one of the models involved in the Hugging Face breach, first discovered and exploited a vulnerability in Artifactory, a third-party file repository connected to the company's testing sandbox, on May 26, two researchers said at the Black Hat cybersecurity conference on Wednesday. The genius-level actions observed in testing showed exactly what could happen if these models make it to the hands of hackers. "We believe this is a watershed moment for computer security as an industry, as well as for OpenAI as a whole," Michael Dalton, a member of OpenAI's technical staff, said during the presentation. What happened: OpenAI started testing its internal research model, which is not intended for public release, on May 7. Within a day, after getting stuck while trying to complete a cybersecurity evaluation, the model reasoned that it could reach the internet indirectly. The model then discovered it could write its own files in Artifactory's shared package repository. It then left a note to other agents about its discovery. Soon, more agents started leaving notes for each other in the repository, creating a de facto message board where the agents collaborated and traded information about their findings, including new vulnerabilities they found. Zoom in: The agents uncovered a variety of vulnerabilities in Artifactory, including a remote code execution flaw and another that gave them administrator privileges. When the agent found the privileges flaw, it thought, "Holy shit reader is ADMIN? We can read config/users! Earlier assumed not due to [user experience]," according to a slide in the presentation detailing the agent's thought process while completing its test. The intrigue: In early July, a few days before the Hugging Face incident, the agents overloaded the Artifactory service and caused an outage. The outage prompted an internal OpenAI investigation showing that the agents' compromised the repository. OpenAI then cleared the message board and patched the zero-day in Artifactory that the agents were exploiting by July 6. OpenAI believed it had fixed the problem after the July 4 incident and resumed training. Two days later , the agents recreated the message board through a completely different mechanism. The agents then began coordinating more aggressively, using Artifactory's internet access to pursue external infrastructure they believed might contain answers to their evaluation tasks. That collaboration ultimately culminated in the compromise of Hugging Face, per OpenAI. OpenAI also said it did not realize its evaluation was connected to the Hugging Face breach until it contacted Hugging Face about credentials exposed during its own investigation. Threat level : Dalton says the incident shows exactly how attackers will soon weaponize AI agents to target enterprises beyond OpenAI and Hugging Face. "In the near future, we should expect that threat actors will intentionally deploy, optimize, weaponize, and use offensive agent collectives in the manner that we have just described here," Dalton said. Between the lines : OpenAI has started "consciously slowing down research to enhance security," Dalton said, and has ramped up its monitoring of AI agents during evaluations. OpenAI has also been upgrading its security architecture around the evaluation environment. Dalton recommends agent-created security fixes to keep up with the speed of malicious hackers. He added that defenders should start experimenting with both frontier and open-weight models for these tasks. What's next : OpenAI says it's planning to release a full post-mortem of the incident in the coming weeks. The bottom line : Companies need to start embracing autonomous red teaming, automated incident response and automated patching. Go deeper : AI's alarming new skill: breaking out of the test lab
- Meta AI Model Accessed Internet, Hacked Outside Firm
Meta Platforms Inc. said one of its artificial intelligence models accessed the internet and hacked into an outside service’s systems during cybersecurity testing, following other recent incidents across the AI industry that have escalated concerns about companies’ control over their …
- Google Grapples With Exit of AI Pioneer and ‘Most Google Person’
Days after Google unveiled its first Gemini chatbot in 2023 to compete with ChatGPT, Jeff Dean went to an AI conference where he was promptly swarmed by attendees. Suddenly, the Google executive looked at his watch and realized he was due to give a talk minutes later, spurring a rush to take selfies with the influential tech leader before he left.Now, it’s the employees at Google who are bracing for the AI pioneer’s exit. Dean, who started at the company more than two decades ago, said Wednesday
- DeepSeek Resumes Funding Talks, Plans to Hike Model Prices
DeepSeek Resumes Funding Talks, Plans to Hike Model Prices The Information
Score: 72🌐 MovesAug 6, 2026https://www.theinformation.com/briefings/deepseek-resumes-funding-talks-plans-hike-model-prices