AI Daily: Unauthorized users access Anthropic’s powerful Mythos model

Catch up on the top artificial intelligence news and commentary by Wall Street analysts on publicly traded companies in the space with this daily recap compiled by The Fly.

Claim 30% Off TipRanks

UNAUTHORIZED USERS: A handful of unauthorized users have accessed Anthropic’s new Mythos AI model, a technology the company claims is so powerful it can enable dangerous cyberattacks, Bloomberg’s Rachel Metz reports, citing to a person familiar with the matter and documentation viewed by Bloomberg News. A small group of users in a private online forum gained access to the model on the same day that the company first announce a plan to release the model to a limited number of companies for testing, and has since been using Mythos regularly, though not for cybersecurity purposes, the author notes.

ANTHROPIC DEAL: President Donald Trump told CNBC that a deal with Anthropic for the Department of Defense to use its artificial intelligence models is “possible,” CNBC’s Ashley Capoot reports. The Pentagon labeled the AI startup a supply chain risk at the beginning of March after negotiations broke down. Trump said his administration had “some very good talks” with Anthropic during a meeting at the White House last week.

WORKSPACE AGENTS: OpenAI is introducing workspace agents in ChatGPT. Teams can now create shared agents that handle complex tasks and long-running workflows, all while operating within the permissions and controls set by their organization. “Workspace agents are an evolution of GPTs. Powered by Codex, they can take on many of the tasks people already do at work-from preparing reports, to writing code, to responding to messages. They run in the cloud, so they can keep working even when you’re not. They’re also designed to be shared within an organization, so teams can build an agent once, use it together in ChatGPT or Slack, and improve it over time,” the company explained. Workspace agents can gather context from the right systems, follow team processes, ask for approval when needed, and keep work moving across tools. Workspace agents are available in research preview in ChatGPT Business, Enterprise, Edu, and Teachers plans. GPTs will remain available while teams test workspace agents with their workflows. “Soon, we’ll make it easy to convert GPTs into workspace agents,” added OpenAI.

INVESTING IN OPENAI: Robinhood’s (HOOD) venture fund is investing $75M in OpenAI, signaling a reconciliation after last year’s dispute over Robinhood’s tokenized “access” to OpenAI shares, which OpenAI publicly disavowed, The New York Times’ Niko Gallogly reports. The deal reflects growing retail investor demand for private tech companies, as OpenAI itself continues expanding ways for public investors to gain exposure through fundraising and ETF partnerships, Gallogy writes.

CRIMINAL PROBE: Florida Attorney General James Uthmeier announced the Office of Statewide Prosecution launched a criminal investigation into OpenAI and its artificial intelligence app, ChatGPT. The decision to launch the investigation comes after an initial review by prosecutors of the chat logs between ChatGPT and the gunman who opened fire at Florida State University last year, Phoenix Ikner. “Florida is leading the way in cracking down on AI’s use in criminal behavior, and if ChatGPT were a person, it would be facing charges for murder,” said Attorney General James Uthmeier. “This criminal investigation will determine whether OpenAI bears criminal responsibility for ChatGPT’s actions in the shooting at Florida State University last year.”

ADVANCED AI MODELS: “We are at an inflection point in cybersecurity,” Microsoft (MSFT) said in a blog post. “Recent advances in AI model capabilities are changing how vulnerabilities are discovered and exploited. AI models can autonomously discover weaknesses, chain multiple lower-severity issues into working end-to-end exploits, and produce working proof-of-concept code. This significantly compresses the window between vulnerability discovery and exploitation…Through Project Glasswing, Microsoft is working closely with Anthropic and industry partners to test Claude Mythos Preview, identify and mitigate vulnerabilities earlier, and coordinate defensive response. We evaluated Mythos using CTI-REALM, our open-source benchmark for real-world detection engineering tasks, and the results showed substantial improvements relative to prior models. Microsoft is also evaluating other models. As part of our overall security approach, we continuously evaluate models from multiple providers as they are made available and integrate them into our enterprise-grade security platform. This multi-model approach is intentional as no single model defines our strategy,” the company explained. “We plan to incorporate advanced AI models, like Claude Mythos Preview, directly into our Security Development Lifecycle (SDL) to identify vulnerabilities and develop mitigations and updates… Beyond plans to use advanced AI models directly into our Security Development Lifecycle (SDL), we are separately building new solutions to help customers leverage advanced AI models to improve their security at enterprise scale.”

EIGHTH GENERATION TPUS: In a blog post, Google (GOOGL) writes, “Today at Google Cloud Next, we are introducing the eighth generation of Google’s custom Tensor Processor Unit, coming soon with two distinct, purpose-built architectures for training and inference: TPU 8t and TPU 8i. These two chips are designed to power our custom-built supercomputers, to drive everything from cutting-edge model training and agent development, to massive inference workloads. TPUs have been powering leading foundation models, including Gemini, for years. These 8th generation TPUs together will deliver scale, efficiency and capabilities across training, serving and agentic workloads. In this age of AI agents, models must reason through problems, execute multi-step workflows and learn from their own actions in continuous loops. This places a new set of demands on infrastructure, and TPU 8t and TPU 8i were designed in partnership with Google DeepMind to take on the most demanding AI workloads and adapt to evolving model architectures at scale.”

MACHINE-LEARNING COMPUTE SPEND: The pace of technological change since last year’s Cloud Next “has never been faster, and Google Cloud has incredible momentum,” Sundar Pichai, CEO of Google and Alphabet, said in a blog post. “Our first-party models now process more than 16 billion tokens per minute via direct API use by our customers, up from 10 billion last quarter. To support and drive this growth, in 2026, just over half of our overall machine learning compute investment is expected to go towards the Cloud business to benefit our cloud customers and partners.” The executive highlighted four key areas: agentic Gemini era; use of AI to defend against security threats; eighth-generation TPUs; and staying on the cutting-edge as “customer zero.”

BUY SHARONAI: Compass Point analyst Michael Donovan initiated coverage of SharonAI (SHAZ) with a Buy rating and $50 price target. With primary operations in Australia, the company focuses on AI and high-performance computing infrastructure, the analyst tells investors in a research note. The firm believes SharonAI’s first major contract moves it beyond setup and into contract-backed scale. In addition, capacity build in Australia gives the company a “credible base from which to deploy,” contends Compass.

ENTERPRISE AI COLLABORATION: Infosys (INFY) announced a strategic collaboration with OpenAI to help enterprises transform software development and modernization with OpenAI’s AI models and products like Codex. Through this collaboration, Infosys will combine OpenAI’s technology with Infosys Topaz Fabric, its agentic services suite. Salil Parekh, CEO, Infosys, said, “Generative and Agentic AI will redefine how enterprises operate and grow. Our collaboration with OpenAI establishes an operating model to unlock AI value at scale – uniting technology, talent, and transformation playbooks so clients can move decisively from pilots to performance, creating competitive advantage. Together, we are not just shaping the future of AI adoption but also enabling our clients to lead it with purpose.”