AI Pulse: Daily Digest — April 22, 2026

Ars Technica

Apple's longtime hardware chief John Ternus will succeed Tim Cook as CEO, with Cook moving to an executive chairman role. The transition hands Ternus the reins of one of the world's most valuable companies at a critical moment — Apple has notably lagged competitors in AI, and Siri's stalled evolution remains an urgent unresolved challenge for the incoming chief.

TechCrunch AI

Amazon has deepened its bet on Anthropic with a fresh $5 billion investment, and Anthropic has committed to spending $100 billion on AWS infrastructure in return — including a massive allocation of Amazon's custom AI chips. The circular deal underscores how tightly the two companies' fates are now intertwined as demand for Claude continues to surge.

TechCrunch AI

Reports claim an unauthorized group has obtained access to Mythos, Anthropic's tightly controlled AI cybersecurity model that was previously known to be in use by the NSA. Anthropic says it is investigating but has found no evidence its core systems were compromised — a statement that will do little to quiet concerns about the proliferation of powerful offensive-capable AI tools.

Ars Technica

Mozilla's CTO revealed that Anthropic's Mythos model discovered 271 previously unknown security vulnerabilities in Firefox 150, calling the AI "every bit as capable" as the world's top human security researchers. The disclosure dramatically illustrates both the power and the danger of Mythos — and adds context to why its unauthorized access is so alarming.

TechCrunch AI

OpenAI CEO Sam Altman publicly criticized Anthropic's Mythos cybersecurity model during a podcast appearance, accusing the company of using fear-driven messaging to inflate perceptions of the tool's capabilities. The broadside comes as the two companies compete fiercely for enterprise and government AI contracts, and as Mythos faces simultaneous scrutiny over its unauthorized access.

TechCrunch AI

SpaceX has struck a deal giving it the option to acquire Cursor, the popular AI coding assistant, for $60 billion — or pay a $10 billion fee if it walks away. The arrangement is widely seen as a strategic move to bolster xAI's developer tools against Anthropic and OpenAI, though analysts note that neither Cursor nor xAI currently has proprietary models competitive with the market leaders.

TechCrunch AI

Meta has deployed an internal surveillance tool that captures employee mouse movements, clicks, and keystrokes to generate training data for its AI agent models. The move highlights the industry's growing hunger for high-quality interactive behavioral data — and raises pointed questions about employee privacy and consent even within a corporate context.

Ars Technica

Florida authorities are investigating whether ChatGPT played a role in a mass shooting, marking one of the most serious government probes yet into AI's potential complicity in real-world violence. OpenAI has stated the chatbot bears no responsibility, but the investigation is likely to intensify regulatory and public scrutiny of AI safety guardrails.

Ars Technica

The U.S. Department of Defense has proposed a $54 billion investment in drone technology, a figure that exceeds the total military budgets of most countries and rivals Ukraine's entire defense spend. The request signals a fundamental shift in how the Pentagon envisions future warfare, with autonomous and AI-guided systems at the center of its strategy.

TechCrunch AI

OpenAI has rolled out ChatGPT Images 2.0, a significantly upgraded image generation model that can search the web in real time to inform its outputs and demonstrates markedly improved ability to render accurate text within images. The update positions OpenAI's image tools more directly against specialized competitors while deepening ChatGPT's role as an all-in-one creative platform.

The Verge

Public anxiety about AI — from data center construction in local communities to job displacement fears — is beginning to surface as a meaningful political issue heading into the 2026 midterms. While most campaigns haven't made AI a centerpiece issue, grassroots resistance to AI infrastructure projects and growing social media anger at tech executives suggest the political reckoning for the industry may be arriving faster than expected.

TechCrunch AI

AI company Clarifai has deleted approximately 3 million photos it obtained from dating platform OkCupid to train facial recognition models, following a settlement with the FTC. Court documents reveal the data-sharing arrangement dates to 2014 and was entangled with personal investment ties between OkCupid executives and Clarifai — a conflict of interest now at the center of regulatory action.

TechCrunch AI

YouTube is rolling out its AI-powered likeness detection tool to celebrities and their representatives, enabling them to automatically scan the platform for deepfake content and request its removal. The expansion reflects mounting pressure on platforms to give public figures meaningful recourse against non-consensual AI-generated media.

TechCrunch AI

AI research startup NeoCognition has closed a $40 million seed round to develop agents capable of acquiring expertise in any domain through human-like learning processes rather than static pre-training. Founded by an Ohio State University researcher, the company is betting that more adaptive, continually learning agents represent the next frontier beyond today's fixed-weight models.

TechCrunch AI

Fermi, the AI-focused nuclear power startup co-founded by former U.S. Energy Secretary Rick Perry, has seen its CEO and CFO depart simultaneously as the company faces significant headwinds with its planned AI data center campus in Texas. The sudden dual departure raises serious questions about the viability of the project at a time when the race to secure reliable power for AI infrastructure is intensifying across the industry.