California Governor Gavin Newsom has just signed an executive order considered the most powerful step ever to control artificial intelligence (AI) at the state level.
The decree requires all technology companies wishing to sign contracts with state authorities to meet strict standards of transparency and safety.
This is seen as part of efforts to make California a pioneer in building a legal framework for AI, a rapidly developing but potentially risky field.
One of the notable points is the regulation on checking the background and operating policies of AI contractors.
Companies must publicize how their systems process data, prevent abuse, especially dangerous content such as child abuse materials.
In addition, California will also carefully assess whether AI technologies, including chatbots and language models, are used to monitor people or censor speech.
Businesses are forced to prove that their systems minimize bias and ensure fairness in the operation process.
The California decree also emphasizes the right to independent evaluation for contractors, even if the federal government issues a risk warning.
This is particularly noteworthy in the context of disputes between the Pentagon and AI Anthropic.
Previously, the Pentagon terminated a contract with Anthropic after the company refused to allow its technology to be used for large-scale domestic surveillance and automated weapons.
However, according to the new decree, California can re-evaluate itself and still allow cooperation if no similar risks are detected.
This move shows the increasingly clear difference between state and federal governments in their approach to AI, especially related to military applications and surveillance.
Another new point is the requirement to affix watermarks to content created or edited by AI, especially videos.
The goal is to limit the spread of misinformation and help people easily distinguish between real content and content created by machines.
Experts assess that this is a necessary step in the context of deepfake technology becoming increasingly sophisticated, making it difficult to verify information in cyberspace.
