Good morning. It’s Thursday, August 13. Twitch just admitted the quiet part out loud: its AI training opt-out only exists because an opt-in system would get zero takers, and millions of streamers now have to dig through settings to keep their content out of Amazon’s model training pipeline.
Today we are covering:
Twitch streamers are opted in by default to having their content train Amazon’s generative AI models, with Twitch admitting an opt-in system would get zero takers
Trump signed a memo letting private firms run offensive cyber operations against foreign criminals, a shift Bloomberg compares to 19th-century “letters of marque”
A Financial Times analysis finds roughly 53 million Chinese workers have been pushed into the gig economy as property and consumer spending slump
Microsoft confirmed it will merge Copilot and Microsoft 365 Copilot into one unified “super app,” rolling out worldwide starting in August
SpaceXAI launched Grok 4.6, tying OpenAI’s GPT-5.6 Sol for third-best model on the Artificial Analysis Intelligence Index
Let’s dive in
Amazon will train on Twitch streamers’ content by default, unless they opt out
By Amanda Silberling via TechCrunch
Twitch will now use creators’ livestream content to train generative AI models for parent company Amazon, opting users in by default and requiring them to manually turn the setting off if they don’t want their content used.
Speaking to nearly 3,000 viewers on a livestream, Twitch Chief Product Officer Mike Minton said plainly, “If this was opt-in, nobody would opt in. That’s honestly the answer,” after Twitch framed the change as merely “adding a setting” rather than announcing that training was already happening.
Twitch Head of Community Mary Kish noted the platform isn’t unique, pointing to Meta’s similar use of public Facebook and Instagram content; users can opt out via channel settings under security and privacy.
𝕏: This is on by default, by the way. They’re painting it in a more positive light by saying you have the freedom to opt out of SOME of this. Creator platforms really hate their creators, huh. Looking forward to my anti AI spiel on my front page stream ❤️ do better. - ShinyCatherine (@shinyycatherine)
Trump Enlists Private Sector to Boost Cyber-Offensive Arsenal
By Patrick Howell O’Neill via Bloomberg
President Trump signed a National Security Presidential Memorandum authorizing private companies, under federal oversight, to run offensive cyber operations against foreign criminal groups targeting Americans, launching a new program run by the Department of Homeland Security.
The program splits authorized activity into surveillance operations that gather intelligence on foreign systems and cyber effects operations that can disrupt, degrade or destroy foreign infrastructure; participating firms must post a bond of at least $1 million that is forfeited for noncompliance.
Bloomberg likens the arrangement to 19th-century “letters of marque,” the government licenses that once let private ships attack enemies, marking a rare shift toward outsourcing offensive hacking to the private sector.
𝕏: This is a pretty big shift in U.S. cyber policy. The White House is setting up a program that would let private cybersecurity companies conduct government-authorized operations against foreign cybercriminal groups, including surveillance and disruption of their infrastructure. Not exactly “hack back,” but definitely a major expansion of the private sector’s role in offensive cyber operations. - Chris Wysopal (@WeldPond)
The best way to reach new readers is through word of mouth. If you click THIS LINK in your inbox, it’ll create an easy-to-send pre-written email you can just fire off to some friends.
China’s great jobs squeeze
Financial Times
A Financial Times analysis finds roughly 53 million Chinese workers have been pushed into the country’s low-wage gig economy, driven into ridesharing and delivery work as traditional job openings dry up.
A prolonged property bust, a consumer spending downturn and the spectre of deflation have shrunk formal employment, with licensed taxi drivers in Guangzhou reportedly sleeping in their cars overnight while waiting for a single fare.
The scale of the shift shows how deep the strain runs beneath China’s headline growth numbers, with gig platforms absorbing surplus labor that the formal economy can no longer employ.
𝕏: There’s no economic paradise. In China, 320 million workers are stuck in gig, temp, freelance jobs — euphemistically called “flexible employment.” Of that, 53+ million are drivers for food delivery or ridesharing. If they work 10-12 hours a day, they make $1000 a month — before costs for vehicle, gas, insurance, repairs, etc. - S.L. Kanthan (@Kanthan2030)
Microsoft Copilot 365 app unified experience
By Jess Weatherbed via The Verge
Microsoft confirmed it will merge Copilot and Microsoft 365 Copilot into a single unified app, combining chat, coding, Copilot Cowork and Autopilot AI agents under one experience for both personal and work accounts.
CEO Satya Nadella confirmed the plan on Microsoft’s July 29 earnings call; worldwide rollout for mobile and web starts in mid-August, with Windows and Mac apps following in mid-September, all under the Microsoft Copilot name and a refreshed icon.
Personal and work account data stay separated within the single app, as Microsoft, led on this effort by former Snap executive Jacob Andreou, races to build an all-in-one AI “super app.”
𝕏: Microsoft is combining its Copilot consumer and commercial apps. No more two Copilot icons on the taskbar or system tray. The unified app has a new Copilot icon and ditches most of the unique UX/UI of the consumer app, ready for a “super app” - Tom Warren (@tomwarren)
SpaceXAI debuts Grok 4.6, overtaking Kimi K3’s performance and matching GPT-5.6 Sol for world’s third best on Artificial Analysis
By Carl Franzen via VentureBeat
SpaceXAI, formerly known as xAI, launched Grok 4.6, a frontier AI model built for long-running agents, coding and knowledge work, scoring 61 on the Artificial Analysis Intelligence Index.
That score ties OpenAI’s GPT-5.6 Sol and edges out Moonshot AI’s Kimi K3, placing Grok 4.6 tied for third-best on the index behind Anthropic’s Claude Opus 5 and Fable 5, with pricing at $2 per million input tokens and $6 per million output tokens.
Grok 4.6 gained 5 points over Grok 4.5 in just over a month, with standout agentic performance and cost efficiency as SpaceXAI races to keep pace with OpenAI and Anthropic in the frontier model race.
𝕏: Grok 4.6 is objectively #1 when considering intelligence, speed & cost - Elon Musk (@elonmusk)
Trending in AI
Today’s big AI theme: AI is becoming a geopolitical and economic force all at once, from an AI-driven hacking campaign against Taiwan to hundreds of billions in memory chip investment and open models racing to catch the frontier.
Taiwan says it was targeted last month in AI-driven hacking campaign
Inside SK Hynix’s $720 billion bet to build enough memory for AI
Thanks for reading to the bottom and soaking in our Newslit Daily fueled with highlights for your morning.
I hope you found it interesting and, needless to say, if you have any questions or feedback let me know by hitting reply.
Take care and see you tomorrow!
How was today’s email?
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.