The Agent Watch
Briefing · Articles · Tools · About EN FR DE ES 中文 IT PT SV FI DA

Daily Briefing

9 October 2026 · 5 stories

🔥 Top story

1

Manus raises $500M after Beijing blocked Meta's takeover

On Thursday 8 October 2026, the Sino-Singaporean AI agent Manus closed a $500M round, its first since Beijing forced Meta to abandon a planned $2B acquisition back in April. The round is led by Boyu Capital and IDG Capital, with Tencent and the former Sequoia China as participants. Manus reportedly claims an annualised run rate near $500M (not audited) and is now targeting a $4B valuation. For the general public, it is as if an Asian start-up everyone thought was finished by geopolitics suddenly bounced back with local Chinese capital, proof that the nationality of an AI agent still shapes who will write the cheque. Source : https://www.cnbc.com/2026/10/08/manus-fund-raise-meta-muse-tencent.html

2

Rein Security raises $25M to watch AI agents at work

On Thursday 8 October, Rein Security, a New York and Tel Aviv start-up founded in 2024, raised a $25M Series A. The money will fund a sidecar that sits next to every AI agent and records every action in real time, down to the line of code that triggered it, and ties each action back to the business outcome it produced. Customers Dun & Bradstreet, Lemonade, Flex and Swimlane already run the system. Rein reportedly multiplied revenue by eight and customers by five since January. For the general public, it is as if every digital assistant came with a surveillance camera that records not the screen but the actual intent behind every click. Source : https://reinsec.io/blog/series-a/

3

JetBrains open-sources a model built for coding agents

On Thursday 8 October, JetBrains, the maker of IntelliJ and PyCharm, released Mellum2.1, a 12-billion parameter AI model of which 2.5 billion are active per token, under the open-source Apache 2.0 licence. Trained with reinforcement learning across millions of simulated environments, it scores 47 % on SWE-bench Verified, the leader of its size class. It works as a fast sub-agent for bulk tasks, while a larger model handles the decisions. For the general public, it is as if a tools vendor for developers published a free auxiliary engine, capable of taking over the repetitive work while the conductor stays focused on the hard calls. Source : https://blog.jetbrains.com/ai/2026/10/mellum2-1-gets-to-work-a-fast-open-model-for-coding-agents

4

Vesta channels $100B of mortgage credit through swarms of agents

On Thursday 8 October, Vesta, the San Francisco platform that has been automating mortgage origination since 2020, raised a $30M Series B, for a total of $85M. The round is led by Conversion Capital, with three lender customers (Pennymac, New American Funding and Citi Ventures) joining in. Vesta combines humans and swarms of specialised agents across the credit chain. Revenue reportedly grew twelvefold in a year, and more than $100B in loans already flow through its tools every year. For the general public, it is as if the paperwork of a home purchase were handed to a team of digital employees, each with one job, coordinated by a human lead. Source : https://techcrunch.com/2026/10/08/vesta-raises-30m-as-lenders-adopt-ai-agents

5

Trump treats as enemy anyone who still says AI instead of Super Intelligence

One week after signing two executive orders on 29 September that imposed the term Super Intelligence in place of artificial intelligence across all US federal communications, Donald Trump declared on Thursday 8 October that anyone still using the word artificial intelligence was the enemy. The same day, six AI players (OpenAI, Anthropic, Google, Meta, Microsoft and Nvidia) signed the White House Accords on Super Intelligence, voluntary self-restraint pledges with no penalty and no new legal obligation. As Axios noted, the irony is that the White House website itself still uses the banned term on several pages. For the general public, it is as if a government decided to rename cars into motor vehicles and punished anyone who kept saying the old word, except that it writes the old word on its own signs. Source : https://www.politico.com/news/2026/10/08/trump-ai-super-intelligence-renaming-01112404

📡 To watch

Universal Quantum raises over $100M, a UK record

Universal Quantum, a British start-up spun out of the University of Sussex, closed on Thursday what is now the largest Series A ever by a UK quantum vendor. Co-led by DCVC and Firgun Ventures and joined by David Baszucki (Roblox) as an individual investor, the round will fund trapped-ion qubits made using classical semiconductor techniques. Quantum is now distancing itself from AI as a standalone investment category, with more than $3B raised in the vertical since 2024. Source : https://universalquantum.com/knowledge-hub/universal-quantum-raises-over-100m-series-a-to-scale-quantum-computing-globally

Washington moves to ban remote GPU rentals

The US Bureau of Industry and Security is drafting a new export rule that would stop Chinese AI firms from renting Nvidia GPUs remotely, through data centres in Thailand, Singapore, Malaysia or Japan. The draft aims to close the loophole used by DeepSeek and Moonshot, which rent a server rather than buying the chip. The new rule still needs a Senate vote to give the Commerce Department the legal authority it lacks. Source : https://www.winzheng.com/en/article/us-commerce-remote-gpu-ban-china-ai-compute-crackdown-2026

Six AI giants sign an accord with zero penalty

The White House Accord on Super Intelligence gathers OpenAI, Anthropic, Google, Meta, Microsoft and Nvidia around voluntary commitments: self-restraint, independent audits and an oversight board. No enforcement mechanism, unlike California (AB 2713 + AB 1405) or the European Union (AI Act). The contrast is now the official hallmark of an American AI safety built on pledges, against a European AI safety built on constraints. Source : https://www.politico.com/news/2026/10/08/trump-ai-super-intelligence-renaming-01112404

Cybersecurity giants queue up for AI agents

Nikesh Arora (Palo Alto Networks) said in September that some forty companies are now contesting the market for AI agent security, of which at least twenty are well funded. Rein Security ($25M) joins AIUC ($40M) and WSO2 (Agent Manager GA) to form a new solvable vertical. Gartner sizes the global market at $4.8B in 2027 and $7.7B in 2028, a near doubling in eighteen months. Source : https://reinsec.io/blog/series-a/

📊 Trend

On Friday 9 October 2026, the AI agent news cycle is telling the same story from three angles. On one side, China protects its champions: Manus, which nearly got swallowed by Meta, raises $500M from local investors and proves an agent can survive a political veto. On the other, security becomes a profession in its own right: Rein Security traces every agent action back to the line of code, while one analyst forecasts a $7.7B global market by 2028. Finally, finance subcontractors adopt swarms of agents: Vesta channels more than $100B of mortgage credit a year, while JetBrains ships a free model for coding sub-tasks. The rule taking shape is simple: an AI agent that actually acts, on a bank account, on a sensitive server or on a home purchase, will inevitably become a commercial target, a regulatory target, or both at once.