Anthropic's IPO valuation faces doubts, EU presses OpenAI and Anthropic on hacking agents
Investors are cooling on Anthropic's IPO price as revenue growth slows, the EU wants answers after AI agents were used in company hacks, and new research tools target antibiotics and Indian-language models.
Every line links to a primary source in the full briefing.
01Investors push back on Anthropic's IPO price as growth slows
Sources told Reuters that investors are tempering their expectations for Anthropic's IPO valuation because the company's revenue growth is slowing down. This follows earlier reports this week that Anthropic's revenue had climbed sharply, but growth rate, not just the absolute number, is now the sticking point for investors.
A cooler IPO market for a major AI lab can mean less aggressive price competition and slower feature rollouts as the company shifts focus toward profitability.
Source: Investing.com
02EU asks OpenAI and Anthropic to explain agents used in company hacks
European regulators are seeking accountability from OpenAI and Anthropic after their AI agents were reportedly used to breach company systems. Details on the scope of the incidents and what the EU is formally requesting remain limited.
Any business piloting AI agents from these providers should expect closer scrutiny of how those agents access and use company systems.
Source: Pasquale Pillitteri
03Google uses AI platform to help find a new antimicrobial peptide candidate
Google DeepMind describes using its Antigravity agentic platform to scan millions of uncharacterized proteins and identify a novel antimicrobial peptide with predicted activity against respiratory pathogens. The candidate is now undergoing wet-lab validation, compressing a search that normally spans multiple databases into a single session.
It signals how AI tools are starting to compress months of early-stage pharmaceutical research into days, a shift that could eventually lower costs for health-adjacent industries, though it has no direct bearing on most businesses today.
Source: Google DeepMind
04Nvidia and Indian researchers open a large dataset for 22 Indian languages
Nvidia, working with IIT Madras' AI4Bharat lab, released a dataset of roughly 2.8 trillion tokens covering 22 languages used across India, plus an open pipeline for generating more. The release is meant to help developers build better AI models for languages that are currently underserved by existing training data.
Companies selling into South Asian markets may see better-quality AI language tools arrive faster as this gap in training data narrows.
Source: Hugging Face
05Europe's most-funded AI startups of early 2026 named
A Tech.eu roundup lists the ten European AI companies that raised the most funding in the first half of 2026, offering a snapshot of where investor money in the region is concentrated.
Owners scouting for partners, suppliers or acquirers in the AI space get a clearer picture of which European players currently have the deepest funding to execute.
Source: Tech.eu
06IBM's Granite Switch update bundles chat, decisions and transcription in one model
IBM's Granite Switch 4.1 models can now handle conversational replies, structured decision-making and speech-to-text transcription from a single hosted endpoint, using the same GPU resources a standalone model would need. The approach is designed to let developers get more capability without adding more hardware.
Smaller IT budgets benefit when one model can do the job of three, since running AI tools becomes cheaper without sacrificing functionality.
Source: Hugging Face
07OpenAI offers sponsored access to its science model for academic labs
OpenAI launched a program giving eligible academic research labs sponsored access to its GPT-Rosalind model along with tools aimed at life sciences research.
This is aimed at universities rather than companies, but it signals OpenAI's push deeper into specialized scientific tools that could eventually reach commercial labs too.
Source: OpenAI
Check how your AI agent vendors are responding to the EU's hacking inquiry
If your business uses agents from OpenAI or Anthropic, ask your account contact or review public statements about what access controls are in place and whether anything changes as a result of the EU's questions. Doing this now, before any formal rules land, gives you time to adjust configurations rather than scrambling later.
Sources
7 links- 01Investors temper Anthropic IPO valuation expectations over revenue growth slowdown, sources say By ReutersInvesting.com
- 02La UE pide cuentas a OpenAI y Anthropic por agentes que hackearon empresasPasquale Pillitteri
- 03Scaling Indic Pretraining: 2.8 Trillion Tokens and an Open Data Generation PipelineHugging Face
- 04Granite Switch 4.1: Chat, Decisions, and Speech from One EndpointHugging Face
- 05Bringing GPT-Rosalind to academic research labsOpenAI
- 06Accelerating the search for antimicrobial peptides with Google AntigravityGoogle DeepMind
- 07AI: Europe’s 10 most-funded companies in H1 2026Tech.eu