Anthropic warned in an amended listing filing that AI accelerating AI research is getting close, and Marvell announced a $115 billion optical communications acquisition. The KOSPI plunged 3.6% as doubts grew over returns on AI investment, and OpenAI was reported to have cut its fundraising target. We also round up a joint statement on agentic AI from U.K. regulators and Google's blocking of an Android TV botnet.
🤖 Anthropic Warns in Amended IPO Filing That "AI Self-Improvement Is Near"
Anthropic said in an amended listing filing (S-1/A) submitted to the U.S. Securities and Exchange Commission on November 2 that using AI to accelerate its own research and development could outpace the development of safety measures, and that by its internal estimates it could soon reach its Responsible Scaling Policy's definition of "dramatic progress due to AI." That reads as being close to "recursive self-improvement," in which AI improves itself, but the company drew a line, saying it has not designated any current model as having crossed that threshold. The same filing also disclosed third-quarter revenue of about $17 billion, nine-month revenue of $33.2 billion, and infrastructure-related commitments of about $895 billion, explaining that most of its net loss came from accounting factors such as fair-value changes in convertible notes.
💰 Marvell to Acquire Lumentum for $115 Billion in All-Stock Deal… Biggest Tech Deal of the Year
Data infrastructure chipmaker Marvell Technology announced on November 2 that it will acquire optical communications component maker Lumentum in an all-stock deal valued at $115 billion. Lumentum shareholders will receive 4.5 Marvell shares per share, worth about $1,132 based on the October 28 closing price — a premium of about 23.9% — with the combined company owned about 67% by Marvell shareholders and about 33% by Lumentum shareholders. The tie-up targets demand for optical connectivity in AI data centers, with completion targeted for the first half of 2028, but Marvell shares fell about 13% on the announcement day, showing investor concerns over dilution and integration risk.
🇰🇷 KOSPI Plunges 3.6%… Samsung Electronics and SK Hynix Fall Together on AI Spending Doubts
The KOSPI closed on November 3 at 5,648.79, down 210.29 points (3.59%) from the previous session. As the burden of AI infrastructure spending weighed on U.S. stocks, with Oracle plunging, Samsung Electronics (-4.15%, ₩219,500) and SK Hynix (-5.03%, ₩1,400,000) fell sharply, and Naver (-3.65%) and Kakao (-3.03%) dropped alongside them. Mirae Asset Securities said concerns about returns on AI investment persist, and foreign and institutional selling dragged the index down from early in the session.
💵 OpenAI Reportedly Cuts Fundraising Target from $30 Billion to $20 Billion… Courting UAE Investors
OpenAI has lowered its fundraising target from an initial plan of $30 billion to raising more than $20 billion at a valuation of about $1.25 trillion, according to the Financial Times. A consortium of investors from the United Arab Emirates, including MGX, the Abu Dhabi sovereign-backed AI investment fund, is expected to be the largest participant, and the first close could fall short of the target as some investors trim their commitments. The company has not officially confirmed this, and the raise is seen as securing funds to expand infrastructure ahead of Anthropic's listing.
🏛️ Four U.K. Regulators Issue Joint Statement on Agentic AI… "Existing Law Already Applies"
The Digital Regulation Cooperation Forum (DRCF) — the Competition and Markets Authority (CMA), the Financial Conduct Authority (FCA), the Information Commissioner's Office (ICO), and the Office of Communications (Ofcom) — said in a joint statement on October 30 that existing competition, consumer, data protection, and online safety laws and rules already apply to personal AI agents. It made clear that even as agents grow more autonomous, that does not reduce the responsibility of the organizations that deploy them, and said each regulator is responding through direct engagement with companies, published guidance, and sandbox support. The ICO is accepting comments on agentic AI until November 20, which may hint at the direction of future guidance.
🔒 Google Blocks 1.3 Million-Device Android TV Botnet… Manufacturer Videostrong Permanently Banned
Google's Threat Intelligence Group announced on November 2 that it had disrupted the command-and-control infrastructure of the "XCPE" botnet, a backdoor planted at the factory stage in more than 1.3 million Android TV set-top boxes. Shenzhen-based Videostrong hid the backdoor by loading malicious code after the first boot, following a passed certification check, and Google believes the infected devices were rented out to criminal groups for ad fraud and other schemes. Google permanently expelled the company from the certified Android device ecosystem, and expects the enforcement action to also cut criminal groups' access to at least 10 million additional devices.