Among the AI news through midday November 5 (Japan time), the standout story was OpenAI's announcement that an autonomous pipeline had found 948 Linux kernel vulnerabilities, verified them, and even generated fixes. Meanwhile, Uber disclosed a more sobering figure: only 4 people booked rides through AI agents during its busiest week. It was a day that laid bare both the expectations around AI agents and the current reality.
OpenAI Automatically Detects 948 Linux Kernel Privilege Escalation Vulnerabilities — and Even Generates Patches
On November 4, OpenAI announced in an article titled "A thousand fewer paths to root" that it had built an autonomous pipeline that discovered and verified 948 local privilege escalation vulnerabilities in the Linux kernel and generated fixes for them.
Local privilege escalation (LPE) vulnerabilities are a class of issues in which an attacker uses an ordinary user account on a device as a foothold to seize the system's highest privileges (root). They are a type of vulnerability that frequently translates into real-world damage, serving as an entry point for malware. Precisely because the Linux kernel runs everything from servers to embedded devices, finding and fixing this kind of vulnerability has traditionally depended on the steady, painstaking work of dedicated maintainers and security researchers.
The key point this time is that an AI pipeline reportedly handled the entire flow — discovery → verification → patch generation. That is drawing attention as a real-world example of AI moving beyond code generation into the practical work of security operations. Note, however, that based on the information available at the time of this roundup, it remains unconfirmed whether the generated patches have been merged into the kernel's official repository, and technical details of the pipeline have not been disclosed.
Uber: "Only 4 People Booked Rides via AI Agents in Our Busiest Week"
According to TechCrunch, Uber revealed that in its busiest week of the previous quarter, only 4 users booked rides through AI agents.
Quite a few companies are building out APIs and payment infrastructure for the "agent economy," where agents transact with one another, but this figure underscores the reality that actual consumer usage is still extremely limited. That said, this is a number from a single operator, Uber, and it does not measure agent-mediated usage across the board. How the industry closes the gap between expectation and current reality — the one the figure "4" so symbolically captures — looks set to become a key talking point for gauging how widely agents will be adopted going forward.
Isomorphic Labs in Talks to Raise ~$5 Billion — One of the Largest Rounds in AI Drug Discovery
Bloomberg reports, citing people familiar with the matter, that Isomorphic Labs — the AI drug discovery company spun out of Alphabet's Google DeepMind — is in discussions to raise roughly $5 billion in new funding. The round is expected to value the company at about $50 billion.
What makes this notable is not just the size of the raise, but the fact that the funds are expected to go toward AI-driven drug discovery (the search for new drugs) — a story that backs the broader trend of generative AI expanding its reach from text and code into the physical territory of the life sciences. That said, the reporting describes negotiations still in progress: the funding is not finalized, and terms could still change.
Ricoh and Kawasaki Heavy Industries Form Strategic Partnership to Leverage Design Knowledge Assets with AI
From Japan, Ricoh and Kawasaki Heavy Industries announced that they have formed a strategic partnership aimed at leveraging knowledge assets from design work through AI.
The two companies plan to use AI to organize and visualize not only the knowledge accumulated in design documents and drawings, but also the "tacit knowledge" of veteran engineers — their experience and know-how — with the aim of supporting technical succession and improving design productivity. As an example of enterprise generative AI evolving from general-purpose tasks like drafting and searching documents toward structuring domain assets that include seasoned engineers' tacit knowledge, the partnership points to one direction for AI adoption in Japanese manufacturing.
Other News
Nothing's CEO said the company's "mysterious AI-first device" will arrive next month. What kind of product it will actually be has not been revealed (TechCrunch).
Euronews covered a report warning that "sovereign AI agents" could evade every kill switch and threaten national security. Concerns over agents operated by state actors are increasingly becoming a subject of international debate.
ITmedia reported that TOTO, best known as a "toilet company," has grown its materials business — which supplies materials used in semiconductor manufacturing equipment — into a pillar that now generates 70% of its operating profit. It is another example of expanding AI demand lifting the entire physical supply chain around it.