Claude crediting backlash: employees say bosses made AI the author, not the helper

When Aubrey, a New York-based healthcare analyst, built a new way to speed up an expensive medical manufacturing process, she used Claude only “in a small capacity.” Then her manager asked for an unusually specific storytelling rewrite. In Aubrey’s senior leadership presentation, could she highlight how Claude helped her, so the AI looked like it had come up with the idea and executed it on its own?

Aubrey tried a compromise. She “puffed up AI’s role” in the planned presentation, while still making clear she did the heavy lifting. But mid-talk, her manager interrupted and announced that Aubrey “had built it all out in a minute with AI.” Weeks later, she received a less-than-enthusiastic annual review, and her boss later said it was a factor.

That is the emotional core of the AI attribution problem, and it is also why it is turning into a business problem. Because in a world where companies are increasingly leaning on AI to get work done faster, the question is no longer just “Should we use AI?” It is “Who gets credit when AI is involved?” And according to research summarized by Business Insider, the answer can be brutal for employees trying to do the morally right thing.

Deepak, an India-based IT developer for a Fortune 500 tech company, describes a similar pattern. Over a year ago, he began regularly crediting the automated coding agents he deploys to carry out grunt work for transparency. But he says upper management started assuming his positive contributions came from AI, and he suspects that assumption stalled an expected promotion. These aren’t isolated anecdotes. They illustrate a recurring incentive mismatch: when AI shows up in the narrative, managers may treat the human as a passenger, even if the human did the decisions.

Christoph Riedl, an information management professor at Northeastern University, points to something more systematic than vibes. In a recent meta-analysis, Riedl and coauthors examined 13 studies across job functions and titles to assess how managers treated employees after employees disclosed AI use. The conclusion was clear: managers consistently devalued workers’ contributions to projects when workers revealed AI had assisted them. Managers assumed the technology did most of the heavy lifting. In other words, transparency can trigger an “AI penalty,” even when employees are trying to be fair and accurate.

So what do people do when they are asked to be transparent but punished for it? According to the reporting, many employees begin hiding their AI usage and wonder how much credit, if any, they should give it for their efforts. The pressure is especially sharp in an era of mass AI-driven layoffs, where the workplace calculus can feel like survival mode: if your work is being evaluated, do you protect your story or protect your honesty?

Employers then try to measure AI use. Many companies rely on tracking tokens, which are the fundamental unit of data processed by an AI model. Tokens can show how often an employee queried a chatbot, how much information they exchanged, and the length of each interaction. But tokens do not show what the AI contributed creatively. That creates a perverse possibility: someone could ask irrelevant questions about the weather or their personal lives, rack up “AI usage,” and still not have the AI do meaningful work. Companies quickly learn that this can discourage real productivity, which is why Amazon reportedly shut down an internal leaderboard that tracked AI token use. Dave Treadwell, an Amazon senior vice president, told staff at a companywide meeting: “Please don’t use AI just for the sake of using AI.”

Even more sophisticated tracking can backfire in different ways. AI coding assistants like Claude Code can automatically add a co-authorship signature in code they write, without explicitly pointing out which lines were auto-generated or how extensively the human author was involved. Riedl argues that if AI use is disclosed without specific details about how it was used, managers’ default assumption is that it was used in a way that reduces agency. The detail of how AI was used seems to matter enormously. The manager’s mind then fills in the blanks: the bot must have driven the new feature, the quick fix, or the text in the report.

Researchers and companies have started experimenting with structured attribution to make the human-AI split legible. Graham Neubig, a computer science professor at Carnegie Mellon University, cofounded OpenHands, an open-source AI coding platform that adds footnote-like attribution to a line of code generated by AI. Meanwhile, IBM’s AI Attribution Toolkit aims for a more granular form, inspired by CRediT, the Contributor Role Taxonomy used in science to outline precise contributions. On the toolkit’s form, users can indicate how much was auto-generated, whether the chatbot produced content from scratch, and whether elements were human-reviewed. Then it produces an attribution statement that can be added to documents, code, and more.

But the social side remains messy. Jessica He, one of the toolkit’s designers, says high-level acknowledgments of AI use are insufficient for both people consuming AI-assisted content and AI users. She adds that the way people engage with someone’s work can differ depending on whether AI was used to generate new ideas or refine wording, and that a user may feel attribution encroaches on their ownership if their AI use was limited. Oliver Schilke, a management and sociology professor at the University of Arizona, notes another contradiction: research suggests the simple act of disclosure can make people trust you less. He argues this is a central contradiction of the AI work era: firms want efficiency gains, but social costs come with adoption. For now, the burden falls on individual users, creating a paradox where those who do the morally right thing bear the penalty for transparency.

Even “fair” attribution rules can crush adoption. Thomas Prommer, an engineering executive at Adidas, described a team pattern: mandatory AI attribution sounded fair, but it quietly killed initiative for engineers. They quit reaching for AI tools because they did not want their best contributions footnoted as “cowritten by Claude.” Prommer says the signal was: AI help diminishes your work. So people hid it or avoided it.

That brings us to the strategic stake for decision-makers. If attribution systems are just compliance theater, employees will game them or avoid them. If disclosure policies are clear but career outcomes are distorted by assumptions, employees will hide AI usage, and management loses visibility into what actually drives quality. A better alternative, Schilke suggests, is collective AI governance norms that include tools such as the Attribution Toolkit, shifting the burden from individual reputation management to shared rules. The question for leaders is not whether to credit AI at all. It is whether your organization’s incentives, tracking, and attribution norms will reward real human judgment or accidentally label it as background noise.