Ad
Skip to content
Read full article about: Android gets AI agents that book trips, fill forms, and clean up your texts

With Gemini Intelligence, Google is introducing new AI features for Android that automate multi-step tasks, summarize web content, fill out forms, and turn spoken thoughts into polished text messages. Ahead of Google I/O, the company announced that the features will ship on the Samsung Galaxy S26 and Google Pixel 10 this summer, with other devices like smartwatches, cars, headsets, and laptops following later in the year.

Gemini is then supposed to handle things like booking trips or moving shopping lists from a notes app straight into a shopping cart. In Chrome, the system summarizes web content and fills out complex forms using autofill - though the latter only kicks in when users explicitly turn it on. A new Gboard feature called Rambler takes spoken, unpolished thoughts and turns them into clean text messages, with support for multiple languages at once. And with "Create My Widget," users can build custom widgets just by describing what they want - recipe suggestions, specific weather data, or anything else.

The push is part of Google's effort to close the gap with OpenAI and Anthropic in the AI agent market. At the beginning of May, the company shut down its experimental browser agent Project Mariner and folded its technology into the new Gemini Agent.

Read full article about: Anthropic expands legal AI offerings with new Claude Cowork plugins

Anthropic is making a major push into the legal sector. On Tuesday, the company unveiled twelve new plugins and over 20 MCP connectors for its chatbot Claude, each designed for specific areas of law, including contract law, employment law, and litigation. The plugins combine skills with connectors to external data sources.

Users can also now connect Claude directly to services like Thomson Reuters' CoCounsel Legal, DocuSign, Everlaw, Box, and the AI legal assistant Harvey. The new features are all available through Claude Cowork. Enterprise admins can enable them in their workspace settings.

According to Anthropic's Chief Legal Officer Mark Pike (via Bloomberg), lawyers now use Claude more than almost any other profession. Over 20,000 lawyers signed up for a recent webinar on using Claude, with another session on the way. Back in February, the launch of Anthropic's first legal tools triggered a trillion-dollar drop in legal software stock prices.

That said, Cowork still has known AI security vulnerabilities like prompt injections. For law firms handling sensitive client data, these are serious concerns worth checking before rolling it out.

Read full article about: Google says it stopped a mass cyberattack after AI was used to discover a zero-day exploit

A new report from Google's Threat Intelligence Group (GTIG) details how attackers are using AI at scale for cyberattacks. For the first time, GTIG identified a threat actor who reportedly used AI to discover and weaponize a zero-day vulnerability. Google says it stopped the planned mass attack.

Comparison of vulnerability discovery tools: According to Google, frontier LLMs are especially effective as general-purpose tools, with capabilities that keep growing. | Source: Google Cloud Blog / GTIG

State-backed actors from China and North Korea are also using AI to hunt for vulnerabilities. The report highlights the GitHub project "wooyun-legacy," a Claude plugin with over 85,000 real vulnerability cases from the Chinese platform WooYun, built to help AI models analyze code more effectively. Russia-linked groups are embedding AI-generated obfuscation code in malware: the Android malware PROMPTSPY, for example, uses the Gemini API to control devices autonomously. Criminal groups like "TeamPCP" are also targeting AI supply chains, going after popular open-source packages, Google says.

Google has developed its own AI-based countermeasures, among them Big Sleep and CodeMender. The full report is available here.

Ad
Read full article about: Alphabet's Isomorphic Labs raises $2.1 billion to scale AI drug discovery toward clinical trials

Isomorphic Labs, the AI drug research company led by Alphabet co-founder Demis Hassabis, has closed a $2.1 billion Series B funding round. Thrive Capital led the round, with participation from Alphabet, GV, MGX, Temasek, CapitalG, and the UK Sovereign AI Fund.

The funding will go toward expanding the company's in-house AI platform for drug development (IsoDDE), pushing its pipeline of drug candidates closer to clinical trials, and hiring globally. IsoDDE combines multiple proprietary AI models designed to work across different therapeutic areas and drug classes.

According to Deepmind co-founder Demis Hassabis, the underlying approach has proven itself. The focus now shifts to scaling the technology, with the ultimate "mission to solve all disease," as Hassabis puts it.

The company was founded in London in 2021 and already has partnerships with Novartis, Lilly, and Johnson & Johnson.

Read full article about: "Tokenmaxxing" spreads at Amazon as employees game internal AI leaderboards

Amazon employees are automating unnecessary tasks just to climb internal AI leaderboards.

The in-house tool "MeshClaw" lets employees create AI agents that can trigger code deployments, triage emails, or interact with apps like Slack. But according to the Financial Times, staff are deliberately using the software to artificially inflate their token consumption.

"There is just so much pressure to use these tools," one Amazon employee said to the FT. "Some people are just using MeshClaw to maximise their token usage." The background: Amazon has set targets for more than 80 percent of developers to use AI each week, and earlier this year began tracking token consumption on internal leaderboards.

Officially, the numbers don't factor into performance reviews. But another employee disagrees: "Managers are looking at it. When they track usage it creates perverse incentives and some people are very competitive about it." Meta employees have engaged in similar "tokenmaxxing."

As a metric for actual productivity gains, token consumption is of little use. We took a closer look at the challenges of meaningfully measuring AI-driven productivity in the latest edition of our Frontier Radar.

Comment Source: FT
Ad

Thinking Machines Lab ships its first model and argues interactivity is what OpenAI gets wrong about voice

Mira Murati’s start-up presents its first AI model and aims to free voice AI from the question-and-answer model. The model processes audio, video and text in 200-millisecond chunks in parallel and aims to beat OpenAI’s GPT Realtime 2 and Google’s Gemini Live in terms of interaction quality.

The EU wants to regulate AI but needs OpenAI and Anthropic to let regulators through the door

OpenAI has offered the EU Commission direct access to its new GPT-5.5 Cyber model for security review, with talks already underway. Anthropic is proving harder to pin down: after four to five meetings on its Mythos model, regulators still don’t have access. The gap highlights how dependent Europe’s AI oversight remains on voluntary cooperation from the companies it aims to regulate.

Ad