AI News Roundup: US-China AI Safety Talks, Palo Alto Warns of AI Cyberattacks, Cisco Cuts 4,000 Jobs

US and China Launch AI Safety Protocol Talks After Beijing Summit

Treasury Secretary Scott Bessent announced on Wednesday that Washington and Beijing will begin formal negotiations on artificial intelligence safety protocols, following meetings between President Trump and Chinese leader Xi Jinping in Beijing. Bessent described the US and China as the world’s “two AI superpowers” and said the talks would establish a framework to prevent the most powerful AI models from reaching non-state actors.

Bessent emphasized that the US is negotiating from a position of strength. He told CNBC that the dialogue is possible precisely because America leads in AI development, adding that the conversation would look very different if China held the technological edge. The announcement came as Bessent met with Chinese Vice Premier He Lifeng in South Korea to discuss broader trade and cooperation issues ahead of the main summit.

Palo Alto Networks Warns AI Cyberattacks Will Be the “New Norm” Within Months

Palo Alto Networks issued a stark warning on Tuesday: organizations have just three to five months before attackers broadly gain access to frontier AI cyber capabilities. The company’s CTO Lee Klarich said businesses are running out of time to shore up their software defenses as hackers increasingly exploit vulnerabilities with AI assistance.

The cybersecurity giant put its own products to the test, using advanced models from Anthropic and OpenAI to probe its software. The result: 75 vulnerabilities discovered — more than seven times the amount typically found in a month. Klarich noted that the latest models are “extraordinarily capable” at finding vulnerabilities and converting them into critical exploit paths in near-real-time. Palo Alto is urging organizations to adopt a four-pronged defense: patch faster, reduce internet-facing exposure, deploy automated detection, and integrate AI into security operations.

Cisco Posts Record $15.8B Revenue, Then Cuts 4,000 Jobs for AI

Cisco delivered a blowout third quarter, reporting record revenue of $15.84 billion — a 12% year-over-year jump that easily beat Wall Street estimates. Shares surged 15% after hours. The networking giant has secured $5.3 billion in AI infrastructure orders year to date and raised its full-year AI order forecast from $5 billion to $9 billion.

But the earnings beat came with a catch: CEO Chuck Robbins announced the company is cutting fewer than 4,000 jobs — roughly 5% of its workforce — to redirect investment toward AI. Most layoff notifications begin today, May 14. The restructuring will cost up to $1 billion in pre-tax charges, with about $450 million hitting the books before fiscal year-end. Cisco also raised its AI-related revenue forecast from $3 billion to $4 billion for fiscal 2026.

Google Rebuilds Android Around Gemini Intelligence

Google unveiled Gemini Intelligence, a sweeping overhaul that reimagines Android as an AI-first platform. The system can move across apps, understand on-screen context, and complete multi-step tasks that previously required jumping between services. A Google executive framed the shift bluntly: “We’re transitioning from an operating system to an intelligence system.”

Key features include app automation, web content summarization, smart form filling, and a new tool called Rambler that polishes spoken messages and builds custom widgets using natural language. The rollout starts this summer on the latest Samsung Galaxy and Google Pixel phones, then expands to watches, cars, glasses, and a new laptop category called Googlebook — designed from the ground up for Gemini. The push comes as Google races to entrench its AI layer before Apple’s expected AI reboot later this year.

Anthropic Hits $30 Billion Revenue Run Rate, Eyes IPO

Anthropic CEO Dario Amodei revealed the company has reached a $30 billion annualized revenue run rate — an 80x increase from its $375 million run rate in early 2025. The growth trajectory has been relentless: $9 billion by end of 2025, $14 billion in February 2026, $19 billion in March, and $30 billion in April. For context, Salesforce took roughly 20 years to reach the same milestone.

The company recently closed a $30 billion Series G at a $380 billion post-money valuation, led by GIC and Coatue. But that may be its last private round: Bloomberg reports Anthropic is weighing an IPO as early as October 2026, with Goldman Sachs, JPMorgan, and Morgan Stanley already in discussions. Fresh funding at a potential $900 billion valuation is also being explored to fund its massive compute needs.

China Blocks Meta’s $2B Acquisition of AI Startup Manus

China’s National Development and Reform Commission formally blocked Meta’s planned $2 billion acquisition of Manus, a Singaporean AI agent startup with Chinese roots. The decision marks the first time China has used state-level authority to prohibit an inbound AI acquisition, with officials describing the deal as a “conspiratorial” attempt to hollow out the country’s technology base.

Meta announced the Manus takeover in December 2025, planning to fold its agent technology directly into Meta AI. Beijing launched a formal probe in January, and after months of scrutiny, ordered the deal reversed. Meta has indicated it will comply, for now. The move underscores how AI is becoming a frontline in US-China tech rivalry — and it adds new context to today’s safety protocol talks between the two nations.