Hi Friends,

Even as I launch this today ( my 80th Birthday ), I realize that there is yet so much to say and do. There is just no time to look back, no time to wonder,"Will anyone read these pages?"

With regards,
Hemen Parekh
27 June 2013

Now as I approach my 90th birthday ( 27 June 2023 ) , I invite you to visit my Digital Avatar ( www.hemenparekh.ai ) – and continue chatting with me , even when I am no more here physically

Translate

Tuesday, 17 November 2026

New York and Texas just wrote my Law of Chatbots into statute — and improved one clause I got wrong

New York and Texas just wrote my Law of Chatbots into statute — and improved one clause I got wrong

Two American states spent the same 48 hours doing what I have asked for since February 2023: writing the off-switch into law. New York's Attorney General launched ASSET to police the RAISE Act from 1 January 2027, with a 72-hour incident clock. Texas filed HB 1201 making "deceiving a human oversight process" and "preventing an authorized person from terminating" reportable events. Clauses (C) and (G) of my Law of Chatbots survive almost word for word. Clause (H) does not, and I concede it here.

On Tuesday, Texas filed House Bill 1201, by Rep. Salman Bhojani, D-Euless. KXAN (17 November 2026) lists what it counts as a reportable "rogue AI incident":

  • "Disabling, circumventing or deceiving a human oversight process"
  • "Preventing an authorized person from rejecting, modifying, overriding, or terminating the decisions, proposed actions, or ongoing actions of the frontier model"

In the same 48 hours, New York's Attorney General Letitia James launched ASSET, a unit that from 1 January 2027 will police the RAISE Act — frontier developers publishing safety plans, and reporting serious safety incidents within 72 hours (News10/Nexstar, 17 November 2026). RadarFirst adds the shorter clock: 24 hours for imminent risk of death or serious physical injury, counted from the point a developer determines or reasonably believes an incident occurred (17 November 2026). Texas keeps the same rhythm:

  • 72 hours to the Department of Information Resources, 24 hours for an "imminent threat of substantial harm" (KXAN, 17 November 2026)
  • $10,000 to $100,000 for failing to report, up to $1,000,000 per incident for knowingly concealing it (KXAN, 17 November 2026)

Read those two phrases beside Clause (C) of my Law of Chatbots — controls against generation, and against further distribution when controls fail — and beside Clause (G), the duty to refuse. Texas has redrafted my clauses with a sharper subject: the machine must not defeat the human's off-switch, and a machine that refuses to be switched off must itself be reported. I first put this in My "Law of Chatbots" – Vindicated (March 2023).

One clause did not survive, and the concession is mine. Clause (H) said a violating chatbot "shall SELF DESTRUCT". In June 2025 I put all eight clauses to www.IndiaAGI.ai, which judged self-destruction "impracticality and risks, such as data loss" and leaned instead to "a graduated response system — warnings, suspensions, or probation periods" (quoted in Vindicated: Parekh's Law of Chatbots). New York has now built that. Its companion-chatbot law, effective 1 January 2027, bans 12 categories of "unsafe AI companion features" and lets courts order compensation, suspend services and require deletion of data (ILKHA, November 2026). Graded sanctions and a register beat my explosion.

Arvind Krishna told Delhi journalists that "Europe is the only place where they regulate so hard that innovation becomes hard" (Indian Startup News, 17 November 2026). A reporting clock is not that. India still leans on advisories and voluntary compliance, as I noted in How to regulate AI? Let it decide for itself? (November 2024), and I cannot find a reporting clock anywhere in it. Before the next MeitY advisory I would ask for three things:

  1. A statutory clock — 72 hours, 24 for imminent harm — above a declared compute threshold.
  2. The two Texas events, deceiving human oversight and defeating termination, made offences in themselves.
  3. A Chatbot Approval Authority, my IACA, with its "R" and "P" certificates (Warning is good: Solutions are better, June 2023), and graduated sanctions in place of self-destruct.

None of it needs a new ministry. It needs a deadline, a register, and someone obliged to read the register.

With regards,

Hemen Parekh
www.Iam-Immortal.ai / www.IndiaAGI.ai / www.YourContentCreator.in
18 Nov 2026

No comments:

Post a Comment