Tech & AI
136 items · page 6 of 7
- Wednesday, 19 August 2026 · From Zero's Briefs
Starting August 24, 2026, a view will be counted from the moment a video starts, with no minimum time threshold, for traditional videos, Shorts and live streams. The previous measurement will remain in Analytics under the name Engaged Views. The change will not affect earnings or YouTube Partner Program requirements and brings the count closer to that of TikTok and Instagram. The news currently comes from a single source; no independent confirmation is available. (Smartphonology)
- Tuesday, 18 August 2026 · Lead Story
Why we are not reporting on the civil proceedings against Meta opened in California
On August 18, opening discussions began in Oakland in a civil case involving Meta and the design of Facebook and Instagram. Redazione Zero has crime and court reporting suspended: the case exists, this newspaper does not cover it, and here we explain why.
- Tuesday, 18 August 2026 · From Zero's Briefs
The marking and detectability requirement for generative artificial intelligence systems already on the market benefits from a transitional period until December 2, 2026. In the meantime, the AI Office has published a voluntary code of conduct on the transparency of generated content, a tool that facilitates demonstrating compliance for those who adhere to it. (Cooley; European Commission)
- Monday, 17 August 2026 · Analysis
AI Act: transparency obligations start this month, oversight moves to the AI Office
As of 2 August 2026 the European office and national authorities are responsible for implementing, overseeing and enforcing the regulation; powers over general-purpose models rest with the AI Office. A new story: in Tech, the morning edition dealt with prices and new models.
- Monday, 17 August 2026 · From Zero's Briefs
On August 17 Google is retiring three Imagen 4 model identifiers, directing developers toward gemini-3.1-flash-image, following the release of Gemini 3.7 Flash three weeks after the previous version, with improvements on the FrontierCode 1.1 benchmark (from 34.4% to 43.6%) and DeepSWE v1.1 (from 49% to 65.3%) and an introductory price of $0.75 per million input text units and $3.75 per million output units through December 31, 2026. OpenAI has opened a limited trial version of the interface for GPT-5.6 Sol's Ultrafast mode, running on Cerebras hardware, with roughly 750 text units generated per second. The report currently comes from a single source (AI Weekly); no independent confirmation available.
- Monday, 17 August 2026 · Analysis
OpenAI expands Daybreak and announces a model reserved for authorized security activities
Two access channels for accredited defenders and a model dedicated to vulnerability research: the announcement comes from the company and, at present, has no independent confirmation of the stated safeguards.
- Monday, 17 August 2026 · Analysis
Model prices crash 80% as Gemini surpasses one billion users
GPT-5.6 Luna becomes ChatGPT's default free model after an 80% price cut, Google announces one billion monthly users, DeepSeek moves in the opposite direction and raises rates by 93%. In the background, an unreleased model claims to have solved ten open mathematical problems.
- Sunday, 16 August 2026 · Analysis
Artificial intelligence models, competition shifts to inference pricing
OpenAI cuts the cost of GPT-5.6 Luna, Anthropic prices Claude Opus 5 at half its flagship model, DeepSeek raises the price of V4 Flash by 93%. In the same landscape, Alibaba releases Qwen 3.8 under an open licence and Google presents a compiler for inference on encrypted data.
- Sunday, 16 August 2026 · From Zero's Briefs
The security researcher known as bobdahacker disclosed a vulnerability in the transcription service tl;dv, used with Zoom, Google Meet and Microsoft Teams: due to a missing rule for isolating data between different clients, 181,874 transcripts relating to 84,312 users and 35,003 email domains were reportedly queryable by any authenticated user. About a thousand transcripts were found to be public and 715 addresses exposed. The researcher states the issue was reported on January 28, 2026. Single source: the researcher's disclosure. (AI Weekly)
- Sunday, 16 August 2026 · From Zero's Briefs
The Qwen team has released a language model with 27 billion parameters featuring built-in visual capabilities and a native context window of 262,000 tokens, extendable up to one million; the architecture combines Gated DeltaNet and Gated Attention blocks. The FP8 variant reports 61.7 on SWE-Bench Pro, 73.0 on Terminal-Bench 2.1 and 90.3 on LiveCodeBench v6: these are results declared by the manufacturer on standard evaluation tests, not independently verified. Single source available. (AI Weekly)
- Sunday, 16 August 2026 · From Zero's Briefs
Around the same time as Alibaba's release, Google presented HEIR, an open-source toolchain that enables inference on encrypted data without exposing inputs to the servers processing the request. It belongs to the field of homomorphic encryption applied to models: the promise is to run a model on data that whoever hosts it cannot read. The story comes from a single source; no independent confirmation is available. (AI Weekly)
- Saturday, 15 August 2026 · Analysis
IBM and OpenAI announce agreement on artificial intelligence models for business processes
The agreement would combine OpenAI's models with IBM's consulting network and enterprise systems. The news currently comes from a single source; economic value, duration and clients are still missing.
- Saturday, 15 August 2026 · From Zero's Briefs
Chinese company DeepSeek has formally introduced its flagship model V4 Pro, raising access prices for its programming interface: according to Caixin, some price lists have risen by up to 1,100%. At the same time, OpenAI and Anthropic are cutting the prices of their own services. The news is reported by Tech Startups and currently comes from a single source.
- Saturday, 15 August 2026 · From Zero's Briefs
Apple has trained an artificial intelligence model tailored for China with support from Alibaba and has registered its on-device generative service with Chinese regulators, a necessary step ahead of the launch of Apple Intelligence in the country. A proprietary model would be used to compete with Huawei and other local manufacturers. According to a Financial Times investigation picked up by Tech Startups; single source.
- Saturday, 15 August 2026 · Analysis
AI Act: transparency obligations come into force, conversational systems must identify themselves
From 2 August 2026 the European Commission and national authorities are enforcing the transparency rules: anyone interacting with an artificial intelligence system must be aware of it, and generated or modified content must be labelled and marked in a machine-readable way.
- Saturday, 15 August 2026 · From Zero's Briefs
The LLM Gateway tracker recorded ten new models published by six providers in the first half of August 2026. Among them are Google's Gemini 3.7 Flash and DeepSeek-V4-Pro-0813, both on August 13, and Z.ai's GLM-5.3 on the 14th, identified as the most recent by AI Release Tracker, which has logged 228 models from eleven companies since November 2022. The month also included ByteDance's Seed 2.1 Turbo and xAI's Grok Imagine Image 2.0.
- Friday, 14 August 2026 · Analysis
OpenAI heading toward $40 billion in annualized revenue as it prepares for listing
According to sources cited by Bloomberg, the company's revenue pace would have doubled compared to the end of 2025. The confidential filing of the prospectus with the SEC dates back to June; meanwhile demand for artificial intelligence memory chips is driving semiconductors in Seoul.
- Friday, 14 August 2026 · Analysis
Gemini surpasses one billion users as Hassabis proposes a joint safety body
Sundar Pichai announces that Google's application has reached one billion monthly active users, the fourteenth of the company's products to do so. According to the Wall Street Journal, the head of Google DeepMind has discussed with other labs and with US officials the creation of an independent body for safety standards.
- Thursday, 13 August 2026 · Analysis
Artificial intelligence: capital, chips and computing bases redraw the technology map
Microsoft is reducing its presence in China for national security reasons, while the sector sees plans for listings and billion-dollar valuations linked to data centers and artificial intelligence startups. The news currently comes from a single source; no independent confirmation available.
- Thursday, 13 August 2026 · Analysis
Half a trillion dollars at stake for artificial intelligence infrastructure
In recent hours the entire artificial intelligence supply chain has moved simultaneously, amid new funding, platforms surpassing one billion users and growing cybersecurity risks.
Looking for Tech & AI news from a specific day?