| View in browser |
|
Triangulate RADAR
Scanning the tech policy landscape this week
|
CERT-In’s AI audit rules put a six-hour clock on AI incidents
One lead, then five quick takes, each with a verdict.
|
|
The lead Prepare
For: AI builders and deployers, Founders, Security, Legal, Investors
- On 4 November, CERT-In issued version 1.0 of its AI Security Auditing Guidelines. They cover AI systems developed, procured, deployed or operated in India, including systems that process Indian data or serve Indian users. Research AI and personal consumer use are left out.
- Every AI system is graded Low, Moderate, High or Critical. The grade turns on how severe an incident could be, how autonomous the model or agent is, how sensitive and exposed the data is, and whether the harm can be reversed. A higher grade brings deeper, more frequent and more independent audits, more documentation and more human oversight.
- Reportable AI security incidents go to CERT-In within six hours, the same clock as the 2022 Directions. Logs are kept for 180 days, with synchronised time across AI components. Our read: these are guidelines, not a statute, but the incident and log duties they cite already bind you for specified incidents under the 2022 Directions. Expect auditors and enterprise buyers to use the tiers as their yardstick.
What to do: List every AI system you build, buy or run, agents included. Grade each from Low to Critical, check that your logs and six-hour escalation path cover them, and name who owns human oversight.
|
|
|
Also on the Radar
| Act now |
Payments. RBI issued a consolidated cyber security master direction on 5 November for non-bank payment operators. It applies at once to large and medium firms; small ones have until 1 April 2028. Expect a 12-hour cooling period after a mobile or email change, a block while remote-access apps run, device binding and six-hour incident reporting to RBI. (MediaNama) |
| Act now |
Consumer apps. The CCPA fined Ola Rs 10 lakh for a pre-ride “tip” that nudged riders to pay before a driver accepted, called it a dark pattern and ordered the prompts removed. Payment is due within 15 days. Rapido was fined the same in September, and Uber and Namma Yatri are under investigation. Any paid “boost” shown before a service is confirmed is now a risk. (The Tribune, ETV Bharat) |
| Prepare |
UPI MDR. NPCI deferred the 0.4% charge on specified UPI merchant payments above Rs 2,000, due 15 October, to “immediately after Diwali”. Diwali was 8 November and we have not found a new date. NPCI has told the Supreme Court that merchants cannot pass the charge on to customers. Price it into your merchant contracts now. (Business Standard, The Tribune) |
| Prepare |
DPDP. Rule 4 came into force on 13 November, opening the route for firms to apply to the Data Protection Board as Consent Managers. We could not confirm that the Board has started taking applications. Nothing changes yet at your signup form: the main duties start on 13 May 2027. Use the six months to map your consent flows. (DPDP Rules, Rule 1) |
| Watch |
Digital arrest scams. The Supreme Court asked WhatsApp and other intermediaries to propose measures to stop digital arrest scams. It also directed banks to compensate victims when their lapses enable the fraud. If you run a messaging, payments or banking product, watch what measures the Court ends up asking for. (MediaNama) |
|
|
Deadlines
Awaited |
MeitY’s AI regulation consultation paper, promised “within a month” of 8 October. We have not found it published. |
To be set |
Start date of the UPI MDR, which NPCI says will follow Diwali |
3 Dec 2026 |
Banks to enforce beneficiary whitelisting or maker-checker approval on APIs used by NBFCs and fintechs, per a confidential RBI advisory reported by MediaNama |
3 Feb 2027 |
Existing NBFC clients of bank APIs to be audited by CERT-In-empanelled auditors (same advisory) |
13 May 2027 |
DPDP main obligations apply (Rules 3, 5 to 16, 22 and 23) |
|
|
Send feedback, corrections or compliments.
|
| Know someone who should read this? Forward this email, or send them to triangulatepolicy.com. |
|
|
|