A Famous Novelist Walked Into OpenAI’s Office. Here’s What He Said.
Today’s AI news has a theme running through it: who gets to trust the tools being built, and who gets left out. A celebrated author confronted the people building ChatGPT. Developers are tightening security around AI software pipelines. And governments are quietly testing AI in one of healthcare’s most frustrating bottlenecks. The stakes are getting more personal.
Dave Eggers Told OpenAI Staff Their Product Was Hurting Writers
The author of A Heartbreaking Work of Staggering Genius reportedly walked into OpenAI’s offices last year — not as a partner or investor, but as a critic. According to The Verge, Dave Eggers spoke directly to staff at Sam Altman’s invitation, arguing that ChatGPT is silencing an entire generation of writers by absorbing the market for creative work.
The concern is specific. When a company pays ChatGPT to write marketing copy, blog posts, or short fiction, that’s a job a human writer didn’t get. Eggers’s argument isn’t just about copyright — it’s about economic displacement. AI models like ChatGPT were trained on vast libraries of human writing, and now they’re competing with the people who produced that writing in the first place.
For anyone who writes for a living — or who wanted to — this is a very real conversation. Freelance rates have already dropped in several creative fields since 2023. Having that conversation happen inside OpenAI’s own walls, with staff listening, is unusual. Whether it changes anything is another question entirely.
Dave Eggers spoke to OpenAI staff about ChatGPT harming writers.
Why this matters: The creative industry isn’t waiting for policy to catch up. Writers are now walking into the rooms where decisions get made.
An Open-Source AI Tool Just Got a Security Upgrade Worth Understanding
LiteLLM — a popular open-source tool that lets developers connect to multiple AI models through a single interface — released two updates this week, versions 1.92.1 and 1.93.0. The headline feature in both: Docker images are now digitally signed using a tool called cosign.
Here’s the plain-English version of what that means. A Docker image is a packaged version of software that runs the same way on any computer. Signing it is like putting a wax seal on a letter — if the seal is broken or missing, you know someone tampered with it. Cosign is the tool that creates and checks that seal. Before this update, someone could theoretically swap out a legitimate LiteLLM image for a malicious one and developers might not notice.
If you’re not a developer, this probably feels abstract. But LiteLLM sits inside a lot of AI-powered products you might use without knowing it. Companies building customer service bots, writing assistants, or internal tools often use libraries like this. When those foundations are more secure, the products built on top of them are more trustworthy.
Docker images signed with cosign for tamper verification.
Why this matters: AI infrastructure security is quietly becoming as important as the models themselves. A compromised pipeline can affect thousands of apps at once.
Also Happening in AI
Chinese AI startup Moonshot AI dropped a new version of its Kimi model this week, with TechCrunch noting the growing competitive pressure it represents in the global AI race. Meanwhile, Google DeepMind introduced DiffusionGemma, a text generation method that produces multiple words simultaneously rather than one at a time — the company claims it’s four times faster than standard approaches, which could matter a lot for real-time applications like voice assistants. On the developer side, Wired reports that Google has changed how it calculates usage limits for its Gemini API, a move that could affect how much it costs to build products on top of Google’s AI. Over in healthcare, Ars Technica is covering a U.S. government pilot using AI to speed up prior authorization — the frustrating process where insurers make patients wait for approval before covering a treatment. And for the curious, a Reddit user built an interactive map of GPT-2’s entire vocabulary space, letting you explore how the model clusters related words and concepts together, which is a genuinely fun way to understand how language models think.
What to Watch
The Eggers story is the visible tip of a much larger conflict between AI companies and creative professionals — one that will almost certainly land in courts and in Congress before the end of 2026. Watch for whether other well-known writers or artists accept similar invitations, and whether OpenAI starts making concrete commitments to the people whose work trained its models. The healthcare AI pilot is worth following just as closely: if AI speeds up prior authorization, it helps patients — but if it just gives insurers a faster way to say no, it makes a broken system worse.