AI Reshapes Search, Security, and the Open Web

Google's AI search is quietly starving the open web of traffic, and this week the consequences became impossible to ignore. From Reddit locking its doors to a landmark data-privacy settlement, the stories below trace a single, uncomfortable thread: AI is redistributing power online, and not always toward users.

Google's AI search is squeezing everyone out

Reddit's CEO told investors this week that Google's AI-powered search mode "does not drive traffic" to the platform. The numbers are stark: search referrals dropped by more than a million visits per day in October alone. Reddit's response is drastic but logical — the company plans to require users to log in to browse the site, effectively ending anonymous access that has existed since Reddit launched. The CEO confirmed the shift on an earnings call, framing it as a way to protect Reddit's audience from leaking out through AI summaries that never send readers anywhere.

If you're learning to build web products, this matters. Designing for logged-in users is a fundamentally different problem than designing for anonymous visitors: authentication flows, personalization, and data retention all become central concerns. The broader pattern is also worth watching — if AI overviews absorb the answer before someone clicks a link, the economic model that funds free online content breaks.

Making things worse for publishers, Google announced it is expanding ads inside AI Overviews, embedding sponsored links directly into follow-up AI queries. Advertisers get visibility; the sites that originally produced the information get less and less.

A major crypto exchange halts withdrawals after a suspected billion-dollar theft

Binance, the world's largest cryptocurrency exchange, paused withdrawals after attackers are suspected of stealing billions in crypto during a cyberattack. The scale, if confirmed, would make it one of the largest crypto heists on record. For anyone studying cybersecurity, this is a reminder that exchanges holding enormous pools of digital assets are high-value, high-complexity targets, and that incident response — including the decision to halt normal operations to limit damage — is itself a skill worth understanding.

DHS employee data exposed in hacktivist breach

A hacker group exposed the personal information of more than 41,000 Department of Homeland Security employees in what appears to be a politically motivated breach targeting government surveillance programs. Separately but thematically connected, fan-fiction site Archive of Our Own (AO3) suffered its own security incident so severe that the platform could remain offline for up to four weeks while staff investigate and recover. The AO3 incident underlines something often overlooked: community-run, volunteer-maintained platforms carry real technical debt, and security incidents can be far more costly for them than for well-funded companies.

GM pays $220 million for selling driver data without consent

General Motors agreed to a $220 million settlement after a coalition of 36 state attorneys general found the company had been selling drivers' location and driving-habit data to third-party brokers without explicit permission. As part of the deal, GM must obtain express consent before collecting or sharing that data going forward. For anyone interested in data science or building connected products, this case is a clear illustration of why privacy-by-design matters: retrofitting consent mechanisms onto a data pipeline that was never built with them is expensive, legally risky, and bad for users.

AWS targets $1 trillion in revenue — if it can outspend rivals first

Amazon CEO Andy Jassy made a bold claim this week: AWS will eventually become the first business in history to hit $1 trillion in annual revenue. The catch is that reaching that number requires continuing a massive capital-expenditure push into AI infrastructure. Amazon's broader quarterly results were mixed — cloud revenue boomed, but the company posted negative free cash flow, a signal that the AI infrastructure race is genuinely expensive even for the biggest players. If you're considering a cloud or backend engineering path, the sheer scale of investment going into AI compute right now is worth understanding — it's where a lot of the job growth is.

Rivian's R2 is outselling everything else in its lineup

Electric vehicle maker Rivian had a strong quarter. The R2 SUV is now officially the company's best-selling vehicle, pushing Rivian to raise its Wall Street guidance for the second time this year. Revenue came in at around $2.15 billion, up 38% year-over-year. On top of that, Rivian expanded its joint venture with Volkswagen to include an AI assistant, which suggests the software layer inside EVs is becoming as important as the hardware underneath. For anyone interested in embedded systems or automotive software, that direction is worth watching.

Google Gemini can now edit your documents directly in chat

A practical update for productivity: Google Gemini users can now generate, preview, and edit Google Docs, Sheets, PDFs, slide decks, and code files without ever leaving the chat interface. It's available to all Gemini users starting now, with web and mobile support varying by file type. For developers and designers who already live inside Google Workspace, this kind of deep integration changes how you might approach prototyping or documentation workflows. It's also a useful case study in how AI assistants are moving from standalone tools into platforms people already use every day.

A ChatGPT lawsuit raises hard questions about AI medical advice

A lawsuit filed in the US alleges that a 31-year-old woman died after ChatGPT advised her to stop taking prescription medication. The case is still in early legal stages, but it puts a sharp focus on a problem that AI developers, ethicists, and regulators are all struggling with: how do you design a general-purpose AI that refuses to give dangerous medical advice without becoming so cautious it's useless? For anyone going into AI development, this is exactly the kind of edge case that safety and alignment research tries to address.

Thrive Holdings, backed by OpenAI, buys ADT for $15 billion

OpenAI-backed Thrive Holdings announced a $15.3 billion acquisition of ADT, the home security company. The stated goal is bringing AI into home security, though specifics about what that looks like in practice remain vague. What is clear is that AI-adjacent investment is now flowing into sectors well beyond software — physical infrastructure, home hardware, and legacy services are all becoming targets for AI-first reinvention.

The pattern across this week's news is consistent: AI is pushing institutions, platforms, and companies to make choices they'd rather not make. Reddit is locking its doors. GM is paying for a data model it built years ago. Amazon is spending before it earns. The tech decisions being made right now will shape what the web looks and feels like for the next decade.