Gemini 2.0 Flash users have discovered the AI model can effectively remove watermarks from existing images

Google’s latest AI innovation, the Gemini 2.0 Flash model, has ignited a firestorm of debate after users discovered its ability to remove watermarks from images, including those from prominent stock media providers like Getty Images. According to reports doing rounds on social media, the model’s image generation feature — rolled out last week as part of an expansion in Google’s developer-facing AI Studio — can not only strip watermarks but also fill in the resulting gaps with surprising accuracy. While labeled “experimental” and “not for production use,” the tool’s lack of restrictions has raised ethical and legal concerns, especially since it’s free to access.

Social media platforms like X and Reddit have buzzed with user demonstrations of Gemini’s capabilities. One X user, George Arrowsmith, showcased the AI seamlessly erasing a watermark from a stock photo, while several Reddit threads praised its gap-filling prowess, though some noted its struggles with semi-transparent or sprawling watermarks. Unlike competitors such as Anthropic’s Claude 3.7 Sonnet or OpenAI’s GPT-4o — which explicitly refuse such tasks, with Claude labeling watermark removal “unethical and potentially illegal” — Gemini 2.0 Flash operates with few guardrails, amplifying its controversial appeal.

Under U.S. copyright law, removing watermarks without consent is generally illegal, barring rare exceptions, as outlined by legal experts. This has left copyright holders wary, especially given the tool’s potential to undermine protections for photographers, artists, and media companies. Google has yet to comment on the backlash, with inquiries sent outside normal business hours unanswered as of March 16, 2025.

This development echoes broader concerns about AI ethics that have surfaced in recent months. In late 2023, OpenAI faced scrutiny when its DALL-E 3 model was found generating near-exact replicas of copyrighted artwork, prompting lawsuits from artists alleging intellectual property theft. Similarly, Anthropic’s Claude was criticized for inadvertently producing biased outputs in text generation, raising questions about accountability in AI deployment. Midjourney has also faced criticism of its own, and so have social media platforms like Instagram. These incidents highlight a recurring theme: as AI tools grow more powerful, their developers struggle to balance innovation with responsibility.

Gemini’s watermark removal feature adds fuel to an already heated debate about AI’s role in creative industries. While some users celebrate the technology as a democratizing force, others see it as a threat to livelihoods dependent on copyright enforcement. The absence of robust restrictions in Gemini 2.0 Flash contrasts sharply with the cautious approaches of its rivals, positioning Google as an outlier in the race to define AI’s ethical boundaries.

For now, the feature remains accessible in Google’s experimental sandbox, but its implications could ripple far beyond. As AI continues to blur the lines between creation and appropriation, regulators and creators alike may demand stricter oversight. Whether Google will impose limits or double down on its permissive stance remains unclear — but the conversation around AI’s power, and its pitfalls, is only growing louder.