日本語
Sources: [1] [2]
2026-07-20 Evening edition
Evening edition

AI News Daily
2026-07-20

Regulators now specify how AI assistants must behave on the phone, while frontier labs keep shipping on capability and price. A retrospective edition: no newly published item cleared the sourcing bar.

Two stories · Regulation and policy · Model release
Today

Highlights

Regulation and policy Published: 2026-07-16
European Commission issues binding specification measures on Google over AI assistant interoperability on Android
Rival assistants must reach the same core Android capabilities as Google’s own AI, with staged deadlines in 2027 and 2028.
Model release Published: 2026-06-09
Anthropic announces Claude Fable 5 and Claude Mythos 5
Fable 5 is a Mythos-class model with safety tuning for general availability, priced at $10 input and $50 output per million tokens.
Editorial note
Retrospective edition — zero qualifying new items in the past 48 hours
Story 01 · Regulation and policy

Brussels sets the terms for AI assistants on Android

Published: 2026-07-16European Commission · Tier 1 official
  • Under the Digital Markets Act, the Commission has issued binding specification measures directed at Google.
  • Competing AI assistants must reach core Android capabilities — voice activation, in-app actions — on the same terms as Gemini and other Google AI.
  • Measures on search data sharing were issued at the same time.
Why it matters

Platform regulation is setting the competitive terms for AI assistants directly. Anyone shipping or integrating AI features on Android needs these dates and requirements as planning inputs.

2027
Android 18 deadline, 1 August
2028
Android 19 deadline, 1 August — several assistants active at once
Source: https://ec.europa.eu/commission/presscorner/detail/en/ip_26_1634
Story 02 · Model release

Anthropic announces Claude Fable 5 and Claude Mythos 5

Published: 2026-06-09Anthropic · Tier 1 official
  • Claude Fable 5 is a Mythos-class model with safety tuning for general availability; Claude Mythos 5 was announced alongside it.
  • Fable 5 is reported to beat previous generally available models on software engineering, knowledge work, visual understanding and scientific research benchmarks.
  • Pricing: $10 per million input tokens, $50 per million output tokens.
Why it matters

Frontier competition continues on capability and price at once. Organisations running Claude in production should revisit the cost-versus-capability trade-off on a regular cycle.

$10
per million input tokens
$50
per million output tokens
Source: https://www.anthropic.com/news/claude-fable-5-mythos-5
Overview

Three currents shaping the picture

01

Regulation as a design constraint

The EU’s Digital Markets Act now reaches into interoperability between AI assistants. For AI companies, compliance is becoming a precondition of product design rather than a downstream legal matter.

02

A fast model refresh cycle

Frontier labs continue to iterate quickly, and competition persists on both capability and price. Evaluation has to be a standing habit, not a one-off decision.

03

Speed versus verifiable sourcing

No item from the past 48 hours met the sourcing bar with sufficient confidence. The gap between how fast information moves and how strictly it can be verified keeps reopening.

Wrap-up

Three things to remember

What to watch next

How Google implements assistant parity before the Android 18 deadline, what concurrent multi-assistant support actually looks like in Android 19, and whether your own cost-versus-capability review keeps pace with the release cycle.