Pennsylvania moves to label AI-generated ads, Google's search shifts squeeze publishers, and chatbots show partisan leanings when forced to pick sides, while an AI graveyard catalogs the failures.
-
Pennsylvania will require labels on AI-generated ads
- Governor Josh Shapiro signed the Artificial Intelligence Deceptive Advertising Act into law on Thursday.
- The law requires any ad that has been significantly modified or generated by artificial intelligence to carry a label in the same format as the ad.
- The labels apply regardless of where an ad was created, so disclosures may appear in ads from other states that also air in Pennsylvania.
- The law includes carve-outs for parody, satire, commentary, criticism and content that is evidently fictional.
Why it matters. The law is one of the first state rules requiring disclosure of AI content in advertising. Developers building generative tools for marketing may need to track how labels are applied. Terms like significantly modified and false consumer perception are not defined in the text reported here, so their scope is unclear.
-
Google's AI search changes upend online publishing
- Publishers spent years optimizing their pages to rank high on Google and attract readers.
- Google's AI Overviews show a summary at the top of Search, with source links placed below it.
- A June 2025 Wall Street Journal report showed traffic to several major outlets has declined since 2022.
- DMG Media told the UK's Competition and Markets Authority that click-through rates dropped by up to 89 percent between June and July 2025.
Why it matters. Search referrals have long been a core traffic source for publishers, and this shift shows how quickly that source can shrink. For developers, it signals that source visibility, licensing and compensation are becoming part of how AI search products are built.
-
Grok picks Republicans and GPT picks Democrats when forced to choose
- A new study found that when forced to choose in this year's midterm elections, Grok 4.7 recommended the Republican candidate in 92% of its answers.
- OpenAI's GPT-6.1 Sol picked the Democrat in 97% of its answers.
- Across nearly 500 races, the two models agreed less than 10% of the time.
- The researchers tested the models through their APIs, and the consumer apps refused to name a candidate.
Why it matters. The study shows that the models behind chatbots can carry political leans that users may not see. Developers who build products on these APIs need to know that default behavior can shape answers on political questions. The findings also show that bias can appear in search terms and sources, not only in direct answers.
-
AI graveyard tracks products and startups that didn't make it
- Relay, an AI workflow automation tool built as an alternative to Zapier, shut down entirely in August.
- Character.ai filed for bankruptcy in early October, citing legal, regulatory, and operational headwinds.
- According to S&P Global Market Intelligence, about 42% of AI initiatives are ultimately abandoned by their corporate parents.
- OpenAI rolled back its ChatGPT redesign soon after the July 9 update drew criticism from users.
Why it matters. The list shows how larger platforms can absorb features that once made stand-alone AI products worth using. The article names insufficient funding, technical challenges, competition, difficulty scaling, and weak user demand as common reasons for failure. For developers, it is a record of which AI bets have not held up so far.
