All posts
#AI today#AI news#Google·November 17, 2026·AI-assisted

AI today: Google finds AI agents exploiting patches with no human in the loop, Alibaba prices Qwen4-Max (17 Nov 2026)

Google's threat team says attackers fully automated finding and exploiting patched flaws with AI for the first time, Alibaba lists Qwen4-Max with a 1-million-token context, Gemini in Apps Script is generally available, the US issues a rule on remote access to AI chips, Genpact launches Onyx, and Karnataka opens the Bengaluru Tech Summit.

Arthat Team
4 min read
AI today: Google finds AI agents exploiting patches with no human in the loop, Alibaba prices Qwen4-Max (17 Nov 2026)

Google's threat intelligence team says that in the past quarter it saw attackers use AI to find and exploit software flaws with no human in the loop, for the first time. The tool compares code before and after a patch, finds the flaw, writes an exploit and runs it.

Google says attackers automated patch-to-exploit with AI

In its latest AI Threat Tracker, dated 17 November, Google Threat Intelligence Group (GTIG) and Mandiant describe a framework they call exploit_agent, found on an exposed staging server. Google says the operator appeared to be targeting French media, entertainment and education websites. The report also covers "LLMJacking", where attackers break into cloud accounts to use up the victim's computing and AI model quotas: Mandiant says single-day fraudulent charges in some cases topped $100,000, and in the worst cases the total reached $1 million before it was stopped.

Why it matters: Google says chaining these steps can "dramatically reduce the window between patch disclosure and weaponized exploitation." Patch internet-facing web software quickly, and put spending alerts on your cloud and AI accounts.

Alibaba lists Qwen4-Max with a 1-million-token context

Alibaba Cloud's Model Studio lists qwen4-max as a new model on 16 November, after a preview that opened on 11 November through its Token Plan only. The model page gives a context length of 1,000,000 tokens and takes text, image and video as input. In the Singapore region (international scope), it lists $1.5 per million input tokens and $4.5 per million output tokens, plus a 1-million-token free quota valid for 90 days.

Qwen4-Max list prices by region
Alibaba Cloud Model Studio, qwen4-max page, checked 17 Nov 2026.

Why it matters: the page lists a different price and deployment scope for each region, so check which region your data would be processed in before you test it on business documents.

Gemini in Apps Script is generally available

Google says the Gemini side panel in Apps Script is now generally available, so anyone in an organisation can describe a task in plain words and have Gemini write the script, per its Workspace Updates post of 16 November. A gradual rollout began on 12 November and can take up to 15 days to show, on Business Standard and Plus, Enterprise Standard and Plus, the AI Expanded Access add-on, Google AI Pro for Education, and Google AI Pro and Ultra.

Why it matters: Gemini proposes code block by block for you to accept or reject. Admins can limit access by organisational unit or group and watch script activity in the Drive and OAuth audit logs.

The US issues a rule on remote access to AI chips

The US Commerce Department's Bureau of Industry and Security released an interim final rule on 16 November on remote access to advanced chips and servers for what it calls "super intelligence" (SI). BIS says the rule was written by and at the direction of the Treasury Department and the National Security Council, and is meant to stop existing restrictions being circumvented through remote computing in data centres in third countries.

Why it matters: BIS says the rule brings remote access in line with the controls on physically shipping the chips. The release doesn't say which customers or countries are affected, so if you rent GPU capacity in data centres outside the US, watch for guidance from your provider.

Genpact launches Onyx for running AI agents in back-office work

Genpact announced the availability of Genpact Onyx, a platform for building, running and governing AI agents in enterprise operations, on 16 November, after unveiling it at an event in New York on 12 November. Genpact says agents start by proposing actions for a human expert to approve, and take on more of the decision as they prove themselves. Its first packaged use is the Genpact Accounts Payable Suite.

Why it matters: a smaller business can copy this for invoices or follow-ups: the agent drafts and a person approves until it has a track record.

Karnataka opens the Bengaluru Tech Summit, sets aside ₹25 crore to buy from startups

Chief Minister D.K. Shivakumar inaugurated the 29th Bengaluru Tech Summit on 17 November, The South First reports. He said that under a Government First initiative the state would be the first customer for startups, with ₹25 crore set aside to buy their solutions.

Why it matters: startups with products a state department could use now have a stated budget to pitch to, though the report gives no application process yet.

Worth watching: Microsoft's Ignite opening keynote, at 10:00 AM PT on 17 November (11:30 PM IST tonight), and Anthropic's IPO, which CNBC reports is expected to price on the evening of 18 November, US time.

Arthat AI

Working on something similar at your company?

Arthat AI builds these automations for Indian businesses every week. If you want to talk through whether this fits your situation, we'll give you a straight answer — even if the answer is "don't build it."

Ready to make AI part of how your business works?

Book a 30-minute discovery call. We’ll listen, tell you where AI pays off first, and be honest if it doesn’t.

See what we’ve built