YOU COULD ASK FOR A HUMAN REVIEW OF AN IMPORTANT AI-ASSISTED DECISION

The proposed AI Bill of Rights would require a person to be told when a high-impact AI system or automated decision-making process is being used on or about them. The notice would have to explain the system's nature and purpose in plain language.

A person could also ask for the principal factors that influenced an automated decision, although the bill allows limits for trade secrets and other protected information. The person could request a meaningful review by a human being who has the authority and competence to confirm, change or reverse the decision.

This could apply when an AI system helps decide whether someone is hired, promoted, disciplined, granted a loan, admitted to a school or given a government benefit. If an applicant is rejected because a system incorrectly matched a name, address or past record, a human review could provide a way to correct the mistake before it follows the person into another application.

Some protection already exists. The National Privacy Commission says people have a right to be informed about automated decision-making involving their personal data. Privacy rules also restrict decisions with legal effects when they are based solely on automated processing without the person's consent. House Bill No. 10362 would state additional rights to protection from discriminatory automated decisions, safety, an explanation of the principal factors and meaningful human review.

The bill does not yet supply a complete complaint procedure. It does not set one response deadline for every request or identify one office that will resolve every AI complaint. Its definition of high-impact AI is broad, covering systems that may significantly affect rights, safety, livelihood, opportunities, essential services or wellbeing. The proposal does not give a specific list of covered systems. Those details could determine whether human review becomes a useful remedy or another request that takes months to answer.

EMPLOYERS WOULD HAVE TO GIVE SIX MONTHS' NOTICE BEFORE A MAJOR AI CHANGE

An employer planning to use high-impact AI that is reasonably expected to substantially alter, reduce or replace the core functions of an employee would have to give at least six months' written notice.

The notice would have to explain the purpose and scope of the system, the work functions affected, the data it will use and how it may influence assignments, performance evaluation, scheduling, hiring, promotion, discipline, pay or termination. Employers would also have to disclose the available retraining, redeployment and transition assistance.

Employees and registered unions would have to be consulted before deployment. A worker facing termination because of AI automation would have to receive reasonable reskilling or upskilling free of charge. If termination still happens, the worker would receive separation benefits under existing law and transition support to be determined by the Department of Labor and Employment.

The bill would also protect a worker from discipline for errors caused by an employer-mandated AI system when the employer failed to provide adequate training, supervision or safeguards. It would prohibit the use of AI to evade security of tenure, minimum wage, benefits and the right to organize.

For a call-center agent, clerk, designer, accountant or other worker whose daily tasks are being changed by AI, six months could provide time to learn a new role, apply for another position or prepare a household budget. Consultation, however, does not give employees a veto over the company's decision. The bill also leaves terms such as "substantially alter" and "reasonable reskilling" open for later rules or interpretation.

For an employer, a fixed six-month period could delay a response to a new technology or a sudden change in its market. Congress will have to decide which changes are large enough to trigger the rule and how employers prove that the required consultation and training were genuine.

The International Labour Organization estimates that more than one in four Philippine jobs are exposed to generative AI, but exposure means that some tasks can be changed or automated. It does not mean that all 12.7 million exposed jobs will disappear. The ILO places 3.6 percent of jobs in the highest exposure category with an elevated displacement risk. The more common result may be a change in what workers do and how their performance is measured.

SOME DEEPFAKES AND OTHER HARMFUL AI USES COULD LEAD TO PRISON

The bill would prohibit the use of AI for identity theft, fraud, deceptive impersonation, child abuse or exploitation, unlawful surveillance and several forms of deceptive synthetic media.

Synthetic media includes AI-generated or AI-altered audio, video and images. The bill would require clear and conspicuous disclosure when synthetic media materially and deceptively alters or realistically depicts a person, statement, event or circumstance in a way likely to mislead the public.

It would also prohibit deceptive synthetic media intended for fraud, public harm, unlawful manipulation or infringement of rights. A person who willfully or knowingly performs, aids, permits or causes a prohibited act could face imprisonment from six months to six years. A public official convicted under the provision could receive temporary or permanent disqualification from public office, while a convicted foreign national could be deported.

This could cover an AI recording made to sound like a relative asking for emergency money, a fabricated video used to damage a candidate, or a false image used to impersonate a person. The proposal does not criminalize every edited photograph, parody or ordinary use of an AI tool. Criminal liability is tied to the prohibited conduct and the requirement that the act be done willfully or knowingly.

The filed text does not define phrases such as "public harm," "unlawful manipulation" and "other unlawful uses." Those terms could cover a wide range of conduct. Congress will have to specify the acts that can lead to prison and clarify how the provision applies to legitimate satire, art, commentary, journalism and harmless editing.

Existing laws already punish conduct such as fraud, identity misuse, child exploitation and some cybercrimes. The proposed act would place AI-specific conduct and synthetic-media disclosure in one framework. Prosecutors and courts would still need to identify the actual prohibited act and prove the required knowledge and intent.

THE BILL WOULD PUT DICT IN CHARGE OF COORDINATION, BUT MANY REMEDIES ARE STILL UNCLEAR

The Department of Information and Communications Technology would lead implementation with help from a multi-sector advisory council. Government members would include the privacy, labor, science, trade, education, energy, planning and local-government agencies. The council would also include representatives from academe, labor, consumer and digital-rights groups, media, youth and the technology industry.

DICT would prepare a national AI strategy within one year and update it every three years. It could establish standards, promote AI literacy, support research, operate regulatory sandboxes and require impact assessments where authorized. A sandbox allows a new system to be tested on a limited scale under supervision. Participation would not excuse a company from liability for harm.

The bill would also require a national AI literacy program in schools and public education for adults and senior citizens. Materials would have to be available in Filipino and major regional languages. For families encountering fake videos, automated scams or AI-generated schoolwork, basic literacy may provide a more immediate benefit than a technical strategy document.

The proposal allows a person injured by negligent, reckless, discriminatory, deceptive or unlawful AI use to sue the deployer, operator or user under existing laws. A developer or provider may also be included in certain cases, such as when the system was primarily designed for illegal use, the provider knew of the illegal use and failed to act, or the system lacked safeguards proportionate to its risk.

A civil case requires time, legal assistance and filing expenses that many Filipinos may struggle to afford. The bill does not create a single independent AI regulator, a complete administrative complaint process or a schedule of administrative fines for every violation. Existing regulators would keep their powers, which means a complaint might belong with the privacy commission, labor department, another sector regulator, a prosecutor or a court depending on the harm.

The proposed right to be forgotten would require deletion of personal data from training datasets, repositories and operational parameters. The bill allows machine unlearning or model retraining when simple deletion is not enough. That remedy may work differently across systems, especially when a model was trained abroad, the data came from several sources or removing one person's influence requires expensive retraining.

AI DATA CENTERS WOULD HAVE TO REPORT THEIR USE OF WATER AND ELECTRICITY

AI does not exist only on a phone or laptop. Training and operating large systems require data centers filled with computers. Those facilities consume electricity for processing and water or other systems for cooling.

House Bill No. 10362 would require AI-related data centers and infrastructure to follow environmental laws, undergo environmental impact assessment and obtain an environmental compliance certificate when applicable. Operators would have to consult communities affected or displaced by their projects.

They would also have to disclose material annual energy and water use, sustainability measures and plans for renewable energy and efficiency, subject to lawful confidentiality limits. DICT would work with the environment and energy departments on standards.

For a community being considered for a large data-center project, the practical questions are local: how much electricity and water will the facility use, where will those resources come from, whether rates or supply could be affected, what jobs will be created and what environmental risks will remain. Annual disclosure would give residents and regulators figures that can be compared with the promises made before construction.

The bill does not set a numerical ceiling for energy or water use. It also does not specify the disclosure format, the threshold for "material" use or the penalty for an incomplete sustainability report. Those details will decide whether disclosure produces comparable public information or a collection of company statements using different measurements.

The proposal applies more widely than Philippine data centers. It covers public and private persons or organizations that develop, deploy, operate, provide or use AI in the country, as well as AI systems that affect people and institutions in the Philippines. Reaching an overseas provider with no meaningful local presence could still be difficult even if the law says the system is covered.