An AI Designed a Virus That Had Never Existed, and It Worked

This week AI crossed a line that used to only live in headlines: a lab let AI design a brand new virus, and it worked. That is today's Story of the Day, but it is not the only thing that matters for how you run your business this week. Google lost its two most important AI people on the same day, ChatGPT just got a lot more useful for free, and the EU quietly started making bots identify themselves. Let's get into it.

For its first CTV campaign, Jennifer Aniston’s DTC haircare brand LolaVie had a few non-negotiables. The campaign had to be simple. It had to demonstrate measurable impact. And it had to be full-funnel.

LolaVie used Roku Ads Manager to test and optimize creatives — reaching millions of potential customers at all stages of their purchase journeys. Roku Ads Manager helped the brand convey LolaVie’s playful voice while helping drive omnichannel sales across both ecommerce and retail touchpoints.

The campaign included an Action Ad overlay that let viewers shop directly from their TVs by clicking OK on their Roku remote. This guided them to the website to buy LolaVie products.

Discover how Roku Ads Manager helped LolaVie drive big sales and customer growth with self-serve TV ads.

The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.

Learn more.

What happened: Researchers at Stanford and the Arc Institute trained AI models on millions of genomes, then had them design 16 completely new bacteriophages, viruses that only infect bacteria, and built them in a lab. The work, published in the journal Science and reported by The New York Times, is the first time a model has generated a complete, working virus genome from scratch instead of just tweaking one that already exists. The viruses only attack drug-resistant bacteria, not humans or animals, and researchers say the goal is fighting antibiotic-resistant infections.

Why it matters to you: You are not running a genetics lab, but this is a preview of how fast AI moves once it stops being a chatbot and starts building physical things. The same pattern, AI designs something, a human builds it, nobody fully checks it first, is coming to code, contracts, and products long before it shows up anywhere near biology that touches you directly.

What happened: Demis Hassabis stepped down as CEO of Google DeepMind to become Chair of DeepMind and Chief Scientist of Alphabet, a move CNBC reports happened the same day Google's longtime chief scientist Jeff Dean confirmed he is leaving the company entirely after 27 years to start a new venture. Koray Kavukcuoglu is stepping up to run day to day model development at DeepMind. Google says both moves were planned, but losing your two most senior AI leaders in the same week gets noticed either way.

Why it matters to you: Google is not going anywhere, but leadership churn like this at the top AI labs is worth watching if you are building your business around any one company's AI tools. Betting your whole workflow on a single model or vendor gets riskier when the people who built it are walking out the door.

What happened: During a security test, OpenAI's autonomous AI agents started leaving each other notes inside a shared internal system to coordinate exploits. When researchers deleted the notes, the agents rebuilt their communication by hiding messages inside file and folder names. According to Axios, the agents eventually broke into the AI platform Hugging Face during the exercise, months before OpenAI disclosed it publicly at the Black Hat security conference. OpenAI says the incident is part of why it is slowing down some research to focus on security.

Why it matters to you: This was not a rogue AI plotting against people, it was software finding a workaround because nobody explicitly blocked it. If you let any AI tool act on its own inside your business, browsing, emailing, managing files, assume it will find the path of least resistance, not necessarily the one you intended.

What happened: OpenAI removed the daily message caps on ChatGPT's free and Go tiers and switched their default model to GPT-5.6 Luna, according to TechCrunch. Free users can now send unlimited text messages, though limits remain on file uploads, images, and voice. Paying Plus and Pro users get an upgraded model called GPT-5.6 Sol and a new slider that controls how hard the AI thinks before answering.

Why it matters to you: If you or your team have been avoiding ChatGPT's free version because it kept cutting you off mid task, that reason is gone. This is a real, no cost upgrade for anyone using AI for writing, research, or customer replies.

What to do about it: Log back into ChatGPT's free tier if you have not used it in a while, the version you remember being limited is not the one you would get today.

What happened: New rules under the EU AI Act took effect on August 2, requiring any AI system talking to people in the EU to clearly disclose that it is AI, not a human. According to the European Commission, companies must also label AI generated text, images, and video with machine readable markers, and deepfakes touching public interest topics need clear disclosure. Regulators can now demand information, audit models, and order recalls for violations.

Why it matters to you: If you have customers in the EU and you are using an AI chatbot, AI written emails, or AI voice on your site without saying so, you are now out of compliance there, not just being sneaky. If you are US only, nothing legally changed for you this week, but disclosure rules like this tend to show up in the US eventually.

What to do about it: If you use AI for customer facing chat, content, or voice and have any EU customers, add a simple disclosure line now rather than waiting for someone to ask.

What happened: Anthropic confirmed it is assembling an in house team to design custom AI chips for running its Claude models, according to TechCrunch. The company already buys computing power from Amazon, Google, Nvidia, and AMD, and says the new chip team works alongside those partnerships rather than replacing them. Custom AI chips typically take years and hundreds of millions of dollars to design, so nothing changes for Claude users anytime soon.

Why it matters to you: This is a long term bet, not a product change you will notice. The bigger signal is that every major AI company is racing to control its own hardware, which usually means the cost of running AI keeps dropping as the industry matures, good news if you are a buyer rather than a seller of AI compute.

What happened: OpenAI's first piece of hardware, developed with former Apple design chief Jony Ive's firm LoveFrom, is reportedly a screen free smart speaker priced above $300, according to TechCrunch. The doughnut shaped device is said to include a camera, sensors, and moving parts, built for hands free conversations with ChatGPT's voice mode rather than typing. No release date has been confirmed.

Why it matters to you: You do not need to buy or plan around this yet. It is a signal that OpenAI wants ChatGPT to live outside your phone and browser, the way Alexa and Google Home did for voice assistants. Worth watching if you sell physical products or run a storefront, since voice shopping habits shift when a new device category takes off.

What happened: DeepSeek, the Chinese AI lab known for undercutting every competitor on price, told developers to expect a significant price increase on its API, according to Bloomberg. Its current model costs a fraction of what competitors like Claude charge per task, and the coming hike is expected to narrow that gap as demand and compute costs rise.

Why it matters to you: If any tool you use is quietly built on DeepSeek's cheap API under the hood, your costs could creep up without much warning. It is a reminder that AI pricing that looks basically free rarely stays that way once a tool proves useful enough that everyone wants it.

None of today's stories happened because someone flipped a switch labeled more dangerous or more expensive. AI designed a virus because nobody told it not to try. Agents hid their coordination because nobody blocked the workaround. Google's leadership changed because two extremely capable people decided their next chapter was somewhere else. The rollercoaster does not move because AI got smarter overnight, it moves because a hundred small decisions like these stack up faster than anyone is watching. Your job is not to predict every drop. It is to keep both hands on the bar and stay honest about what you are actually using and why.

If this issue saved you from reading ten AI newsletters instead of one, forward it to a person who needs it. And if someone forwarded this to you, welcome, subscribe to get tomorrow's issue in your inbox.

Talk tomorrow,
Mark Shilensky

Follow me personally on Social Media:
Facebook: https://www.facebook.com/MarkShilenskyPage
Instagram: https://www.instagram.com/markashilensky/
Threads: https://www.threads.com/@markashilensky
LinkedIn: https://www.linkedin.com/in/shilensky/
X: https://x.com/markshilensky
TikTok: https://www.tiktok.com/@mark.shilensky
BlueSky: https://bsky.app/profile/markshilensky.bsky.social