INFO 202 Chap.13 AI Policy and Governance
AI Policy and Governance
How the course ends
Week 13 is the concluding week, and it turns from what AI does to how societies try to steer it. The syllabus frames the week around the major debates in AI policy and governance, especially the prominent competing positions on regulation and the value each claims to deliver.
The readings are Misra's 2024 essay on the trouble with SB 1047, a 2024 California bill concerned with frontier AI models, and the MIT AI Risk Initiative. The task is not to pick a winner but to explain what each position is trying to protect and at what cost.
Reading a stance through the value it protects
One useful way to organize the debate is to ask what each camp is guarding.
Advocates of a light touch put speed and national competitiveness first and warn that heavy rules can push development abroad. Supporters of safety and security standards want protections in place before harms spread. Those who favor legislation aimed at frontier models want precaution concentrated where capability is greatest.
A liability approach relies on courts to hold companies accountable after harm, and an antitrust approach worries about power pooling in a handful of firms. Each value is legitimate, which is why the debate persists.
Politics and industry in late 2024
A Financial Times report from the week of the 2024 election describes technology executives hoping for a more permissive approach to mergers, antitrust and AI.
It notes that the incoming president had promised to cancel his predecessor's AI executive order and had earlier issued his own order warning that overly restrictive rules could hamper innovation.
The same report shows how mixed the alignments are: a landmark antitrust case against Google began under the same president, and one of his most prominent backers warns that uncontrolled AI threatens humanity while racing to build it.
Liability and the Section 230 question
A Wall Street Journal opinion column by Andy Kessler asks whether algorithms and generative AI make a company a publisher.
Section 230, part of the 1996 Communications Decency Act, shields services from being treated as publishers of content others supply. Kessler points to an appeals court ruling that treated a recommendation algorithm as the platform's own expression, and he argues lawmakers should extend that logic to chatbots.
Read it as an advocacy piece that shows what a liability stance looks like, not as a statement of current law.
Writing about competing stances
Strong answers name the stance, state its value proposition, identify when and through whom it acts, and then test it against a concrete harm. Keep claims about the SB 1047 bill at the level of its purpose unless your course materials give more.
What this chapter covers
- 01
Week 13 scope: competing stances on AI regulation and their value propositions
- 02
SB 1047 as a 2024 California bill on frontier AI models
- 03
Light-touch, innovation-first arguments
- 04
Safety and security standards set before deployment
- 05
Liability for algorithmic and generated output
- 06
Antitrust scrutiny of Big Tech and AI start-up deals
- 07
Why the stances cut across party lines
- 08
Timing and venue as a way to compare regulatory tools
Worked example · free
Compare two regulatory stances on one AI harm
- 2State the light-touch response and the value it protects.
- 2State the liability response and the value it protects.
- 1Reach a reasoned judgment for this particular harm.
Key terms
- Frontier AI Model
- A term for the most capable general-purpose AI systems at a given time, the category that California's SB 1047 of 2024 was written to address.
- Light-Touch Regulation
- A stance that keeps new rules minimal and acts only when harms are demonstrated, on the view that restrictive rules slow innovation or drive it abroad.
- Section 230
- The 1996 provision of US communications law that prevents online services from being treated as the publisher or speaker of content supplied by others.
- Publisher Liability
- Legal responsibility borne by whoever publishes content; advocates argue AI companies should carry it because their systems, not users, produce the output.
- Antitrust Enforcement
- Government action that reviews mergers and challenges monopoly power to protect competition, applied recently to Big Tech partnerships with AI start-ups.
AI Policy and Governance FAQ
What was California's SB 1047?
SB 1047 was a 2024 California bill concerned with frontier AI models, the most capable systems being built. It is a focal case for the debate over whether rules should be written in advance for the most powerful models. Misra's essay on the bill examines what its critics found troubling, which makes it a useful test of how each regulatory stance weighs precaution against innovation.
Does Section 230 protect AI chatbots?
That is an open argument rather than a settled answer. Section 230 protects services from being treated as the publisher of other people's content, but a chatbot generates its own output. Commentators such as Kessler argue this makes AI companies publishers, pointing to a court ruling that treated a recommendation algorithm as a platform's own expression.
Kessler himself calls on Congress to codify that line, a sign the question was still open when he wrote.
Why do tech companies prefer light-touch AI rules?
Light-touch rules lower compliance costs and let firms move quickly, and supporters frame this as a matter of national competitiveness, especially in the race with China. Reporting around the 2024 election described industry leaders hoping for a more permissive stance on mergers, antitrust and AI. Critics answer that waiting for harm leaves the people affected to bear the cost first.
Are AI policy positions just Democrat versus Republican?
Not neatly. The antitrust case that Google lost in 2024 was filed under a Republican administration, and some Republican figures have praised aggressive antitrust enforcers. Leading technologists can also hold mixed positions, warning about existential risk while racing to build more powerful systems. Comparing stances by the value they protect is more reliable than sorting them by party.
Exam move
Draw a table with one row per regulatory stance and three columns: its value proposition, when it acts, and who acts. Fill it from memory, then check each row against the readings. Next, take one concrete AI harm, such as a chatbot inventing a crime, and write two sentences for each stance on how it would respond. Finish by choosing the stance you find strongest for that harm and defending it in four sentences.
Working through AI Policy and Governance in INFO 202? Sia is AskSia’s AI Information Technology tutor — ask any INFO 202 AI Policy and Governance question and get a clear, step-by-step explanation grounded in how INFO 202 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
