SShortSingh.
Back to feed

FCC Rules Cover AI Voice Calls Under US Telemarketing Law

0
·3 views

The FCC's February 2024 Declaratory Ruling confirmed that the Telephone Consumer Protection Act applies to AI-generated voices, including voice cloning that sounds like a human. Restrictions on artificial or prerecorded voice calls apply to residential lines and certain categories of numbers, such as mobile phones. Calls must open by identifying the responsible business or caller, and calls that sell products may also fall under the FTC's Telemarketing Sales Rule. Calls to businesses for information or bookings differ from mass or sales calls, though many small businesses use mobile numbers. State recording-consent laws also apply, and the article notes that this overview is not legal advice.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Developer finds homepage JavaScript feed hid 324 posts from Google's crawler

A blogger spent a month puzzled that Google was not indexing their stories, even though the sitemap listed them and the pages loaded correctly. The root cause was that the homepage's story feed was injected by JavaScript after load, so the raw HTML Google reads contained no links to any stories. A secondary feed page existed but was itself unlinked from the site, leaving crawlers no path to individual articles. The fix was to place 12 story links directly in the homepage HTML, allowing crawlers to reach stories without executing JavaScript. The author notes the bug was invisible in a normal browser, which runs scripts and displays the finished page.

0
ProgrammingDEV Community ·

AI assistant spent 54 seconds thinking and returned no answer to a simple question

On 6 July 2026, a user asked the Merlin assistant a one-line question about what a structured intelligence assistant is. The model used its entire 2,048-token thinking budget without closing its reasoning block, leaving the answer section empty. After 54 seconds, the user received a typed error instead of a response. The developer responded by disabling thinking on the direct-answer path and adding a fallback that salvages answer-like text from unfinished reasoning, tracked through a metric. The developer argues that systems should explicitly define what happens when a reasoning budget runs out mid-thought.

0
ProgrammingDEV Community ·

Guide to Querying US State Business Registries with Python and Open Data

Every US LLC and corporation is registered with a state, and each state keeps a public record covering legal name, entity ID, type, status, formation date, addresses and registered agent. This information helps with know-your-business checks, confirming vendor or customer legitimacy, and tracking newly formed companies. Because each state runs its own registry with different search pages, field names and ID formats, cross-state lookups are inconsistent. Some states publish their registries as official open data, and Colorado's portal on Socrata allows filtering by name or formation date with a simple query API. Scaling beyond one state requires per-state column mappings, status code lookup tables, grouping of multi-row records, and differing update frequencies. The article is written by the author of a paid Apify tool that searches six states at once.

0
ProgrammingDEV Community ·

Toggling Baked Lamps in Three.js Without Reloading Textures

A developer tested switching baked lighting layers on a garden bench scene in Three.js without reloading textures. Blender Cycles baked the overhead fill, amber lamp, and blue lamp contributions onto a shared second UV atlas, and a shader blended the three linear textures using adjustable weights. The 116-triangle fixture kept the bench and path in a single draw call while the lamps were toggled. The layered textures totaled 72,528 bytes, compared with 83,004 bytes for an independent all-on bake, though the layered version added two cold-load requests. On a desktop with Chrome, WebGL 2, Three.js r187 and an RTX 5070 Ti, 24 switches took a median of 13.9 milliseconds from handler entry to the next render call, excluding GPU execution and display latency.