Instagram Is Coming for Fake AI Accounts — And Politicians Are Starting to Notice

A quiet shift is happening across social media and government offices: people are getting serious about who — or what — is actually talking to them online. Today’s stories share a common thread. Whether it’s a governor calling out AI’s harms, a platform cracking down on fake profiles, or a developer tool adding new safeguards, the pressure to make AI more honest is building fast.

New York’s Governor Wants AI to Be Less Harmful. What Does That Actually Mean?

New York Governor Kathy Hochul has reportedly made AI ethics a visible part of her policy platform heading into an election year, according to The Verge. Her stated concern: AI should be less harmful to people. The specifics of any legislation are still vague, but the political signal is clear — AI safety is now a mainstream campaign issue, not just a Silicon Valley talking point.

Think of it like seatbelt laws. For years, car companies built faster, flashier vehicles while safety standards lagged behind. Eventually, public pressure and political will caught up. Hochul seems to be positioning herself at that inflection point for AI, where the technology has moved faster than the rules meant to govern it.

For everyday people, a politician staking out this territory matters because it suggests regulation could come — and sooner than expected. Voters in New York represent a huge market. If state-level rules pass, companies that want access to those consumers will have to comply, whether they’re based in Albany or San Francisco.

Why this matters: When governors start running on AI ethics, the issue has officially crossed into mainstream politics. What gets said on the campaign trail in 2026 often becomes law by 2028.

“Governor Hochul is taking active positions on major technology issues including AI ethics concerns.”

Instagram Is Making Fake AI Profiles Harder to Build a Following

According to TechCrunch, Instagram is reportedly rolling out new restrictions on AI-generated profiles that don’t disclose their artificial nature. In plain terms: if an account is run by AI and doesn’t say so, Instagram will limit how many people can see its posts, effectively capping its reach before it becomes influential.

Imagine someone putting up a storefront with a realistic mannequin in the window, but the mannequin is actually taking orders and giving advice. That’s roughly what undisclosed AI profiles do — they look like real people, build trust, and interact with users who have no idea they’re talking to software. Instagram’s new approach essentially puts a small sign on that store saying “this isn’t a real person” by reducing its visibility to others.

For regular Instagram users, this is meaningful protection. AI accounts can be used to spread misinformation, manipulate trends, or build fake credibility for products and ideas. Reducing their reach before they accumulate large followings cuts the problem off earlier in the process. You’re less likely to stumble across a highly followed “person” who turns out to be a bot.

Why this matters: Disclosure rules only work if platforms enforce them. Limiting reach for non-compliant accounts gives the rule actual teeth.

“Instagram reduces visibility for undisclosed AI profiles to limit deceptive content spread.”

LiteLLM — an open-source library, meaning free software anyone can inspect and use, that lets developers connect to multiple AI services through a single interface — released version 1.99.0 this week. The headline feature is signed Docker images. A Docker image is essentially a pre-packaged bundle of software. “Signing” one means attaching a cryptographic stamp, like a wax seal on a letter, that proves the package hasn’t been tampered with since the developer created it. Users can verify that seal before installing anything. GitHub hosts the full release notes.

Supply chain attacks — where hackers modify legitimate software before it reaches users — have become a real concern in the developer world. Signing packages is a direct response to that threat. It’s the software equivalent of tamper-evident packaging on medicine.

Most people will never install LiteLLM directly. But many apps and services use tools like it under the hood. When the building blocks of software are more secure, the products built on top of them are safer too.

Why this matters: Security improvements in developer tools ripple outward. Safer infrastructure means fewer compromised apps reaching ordinary users.

“Signed Docker images users can verify for authenticity.”

Also Happening in AI

Instagram is pulling double duty this week — beyond limiting undisclosed AI profiles, The Verge reports the platform is also cracking down on AI accounts that actively impersonate real people, a related but distinct problem. On the research side, a paper published on arXiv describes a four-stage method for identifying which AI model is secretly running behind a product, a tool that could become important for regulators and auditors. For developers curious about image generation, Matthew Berman published a practical guide to using OpenRouter — a platform that provides access to multiple AI image tools through one interface. Meanwhile, a Harvard Law dropout raised $6 million to build Blue Voice, an AI assistant trained on police department records to help officers navigate legal and procedural questions. And in the open-source software world, The Verge reports that Debian, a widely used Linux distribution, has decided to allow AI-generated code rather than ban it outright — requiring contributors to disclose it instead.

What to Watch

The Instagram stories today hint at a broader platform reckoning: disclosure rules are spreading, and enforcement is following. Watch whether other major platforms — TikTok, X, YouTube — adopt similar reach-limiting tactics for undisclosed AI accounts in the next few months, or whether Instagram becomes the outlier. On the political side, Governor Hochul’s positioning is worth tracking closely: if she wins her race while running on AI ethics, expect a wave of similar platforms from governors in other large states heading into 2027.