17 May 2026: ArXiv Cracks Down on AI Authors as ChatGPT Adds Finance Dashboard

The week’s AI story is being told in two voices. One is about who gets paid. The other is about what users actually do tomorrow morning, from a research preprint to a personal-finance dashboard inside ChatGPT.

ArXiv said it will ban authors for a year if they let AI do all the work. The free preprint server, where most of the world’s machine learning and physics research is shared before peer review, is tightening its policy on large language models. Authors who produce papers without meaningful human contribution risk a 12 month posting ban, according to TechCrunch.

ArXiv is not banning AI. It is banning lazy use of it. Reviews of model behaviour, citation accuracy and methodological rigour suggest the platform has been flooded with low-effort, machine-generated submissions, particularly in computer science. The policy treats AI as a tool, not a co-author, and puts the onus back on the human to do the thinking.

For UK researchers and anyone who relies on ArXiv to track frontier work, this matters because it sets a norm. If the largest open preprint server insists on substantive human authorship, journals and funders will likely follow. The era of “the model wrote the paper” looks shorter than some assumed.

OpenAI has rolled out a personal finance experience inside ChatGPT for Pro users in the United States. The new Finances tab lets users connect bank, brokerage and card accounts through the data provider Plaid, and ask grounded questions about spending, subscriptions, net worth and savings goals. It also produces a dashboard view, OpenAI announced.

More than 12,000 financial institutions are supported at launch, including Schwab, Fidelity, Chase, Capital One, Robinhood and American Express. The product follows OpenAI’s April acquisition of the personal finance startup Hiro. UK access is not live yet, and the feature is in preview on web and iOS for the $200 a month Pro tier only.

The signal here is bigger than the feature. ChatGPT is moving from a chat box you visit to a place where your money, calendar and email already live. Anyone weighing free against paid AI subscriptions should note that the paid tiers are now where the most useful integrations land first.

Greg Brockman is reportedly taking charge of product strategy at OpenAI. The co-founder is being placed at the centre of a planned merger between ChatGPT, the developer API and the coding agent Codex, TechCrunch reports. The aim is a single product team rather than three competing roadmaps.

This is the second senior shake-up at OpenAI this quarter and points to a future where ChatGPT, the API and Codex are presented as one surface. For developers that may mean a tighter loop between chat and code execution. For business buyers it should mean fewer SKUs to track. The risk is the usual one, that combining everything slows down each part.

Microsoft’s AI chief Mustafa Suleyman gave 18 months as the timeline for white-collar work to be largely automated. Speaking to Fortune, Suleyman framed the next year and a half as a step change, with knowledge tasks moving from human-led to AI-led at scale. He stopped short of saying jobs would disappear, but he did say the work inside them would.

Predictions like this should be read with care. Suleyman runs the consumer AI division at Microsoft and has a commercial interest in the timeline being short. Still, the direction matches what AI agents can already do today, and it is a useful prompt for any business owner asking what to automate before competitors do.

The TechCrunch essay on the AI gold rush captured the mood at the top of the industry. Citing Menlo Ventures partner Deedy Das, the piece notes that roughly 10,000 staff at OpenAI, Anthropic and Nvidia now have paper wealth above $20 million, while layoffs sweep the rest of the software sector, TechCrunch wrote.

The split matters because it shapes what gets built. When a small concentrated group sets product priorities, mainstream user needs can get squeezed out by lab-grade interests. Watch for which features ship to the most expensive tier first, because that is where the gap shows up most clearly.

Here is everything else worth knowing from today’s AI news.

• Anthropic’s run-rate revenue surpasses $30 billion. Up from roughly $9 billion at end of 2025, with over 1,000 customers spending more than $1 million annually. Source
• Google says hackers are using AI to weaponise zero-day vulnerabilities. The threat intelligence group disrupted an attempt to use AI to plan a mass exploitation operation. Source
• AI is a technology, not a product. John Gruber argues that the framing is what is causing investors to confuse short-term hype with long-term shape. Source
• Engineer argues AI will not make your processes faster. A reminder that adding AI to a broken workflow tends to scale the breakage, not the throughput. Source
• Apple Silicon costs more than OpenRouter for offline inference. A blog post crunches the numbers on running models locally versus paying per token. Source
• AI subscriptions called a ticking time bomb for enterprises. The argument is that per-seat pricing breaks down once agents start doing the seats’ work. Source
• OpenAI agrees to give the EU early access to GPT-5.5-Cyber. A test-before-deploy commitment for a security-focused variant of its latest model. Source
• Digg relaunches as an AI news aggregator. The old social bookmarking brand is back, this time pitched as a feed curated by language models. Source

The next thing to watch is whether OpenAI rolls Finances out to Plus subscribers and to the UK before the end of the quarter. The decision will tell you how confident OpenAI is in its account-linking pipeline and whether Plaid’s UK Open Banking coverage is good enough for a full launch. If Finances stays US-Pro-only into June, that is a tell about regulatory friction, not user demand.

This is a daily news update for informational purposes only. AI products and policies change rapidly. Verify details directly with providers before making decisions. Nothing here is financial or legal advice.

AI Daily is Cristoniq’s afternoon update on developments in artificial intelligence, published every weekday afternoon.