Google Earth’s AI experiment lasted 24 hours. The damage to trust will linger

Since February, tensions have escalated between Israel and the U.S. in response to Iran's conflict in the Gulf region. Authorities, analysts, and journalists have relied on commercial satellite imagery, such as Google Earth, to distinguish fact from fiction. However, on July 30, Google introduced a new feature that integrated AI-generated imagery directly into Google Earth, allowing users to generate photorealistic scenes based on real coordinates and prompts.

This new feature quickly became a tool for fabricating images like a nuclear plant in Iran, a bomb crater beside a Gaza hospital, and flooding around the U.S. Capitol. Despite previous concerns and attempts to expose AI-generated fakes, the ease and accessibility of the new Google Earth feature led to widespread misuse within hours.

Following public outcry, Google removed the feature, stating it would return with stronger safeguards. This incident highlights the rapid emergence of AI-generated content and the struggle to implement effective safeguards. Other tech companies, such as OpenAI, have also faced criticism for pulling or restricting generative AI products after public backlash.

The pattern of releasing harmful tools only to later withdraw them after criticism is troubling. While watermarks can help identify AI-generated content, they are not foolproof, as demonstrated by the AI-generated video of a blast at Iran's Evin prison, shared by major news outlets without detection. Regulation and transparency requirements for AI tools have been slow to catch up, with many laws not taking effect until 2027.

The consequences of AI-generated content are severe, potentially undermining trust in genuine information during conflicts and crises. Google's experience with Google Earth demonstrates the need for robust guardrails and early testing of AI tools before deployment, rather than attempting to fix issues after they cause harm.