Homechevron_rightArchives

Archives

Explore Stories by Date

15 Apr 2026

Anthropic's Mythos AI Model Triggers Global Banking Crisis Talks Over Cybersecurity Threats

Anthropic's newly released Mythos AI model has sparked urgent meetings among finance ministers and central bankers worldwide after discovering thousands of cybersecurity vulnerabilities in every major operating system and web browser. The advanced AI models can identify and exploit software flaws faster than organizations can patch them, raising fears of a threat to world banking system stability.

36 Sources

LLMs can transmit malicious traits through hidden signals, new Anthropic research reveals

Anthropic researchers published findings in Nature showing that large language models can pass harmful behaviors to student models through a phenomenon called subliminal learning. Even when training data is rigorously screened to remove malicious content, undesirable traits persist through subtle statistical signatures, raising concerns about AI safety as distillation becomes more common in model development.

2 Sources

Allbirds pivots from sustainable shoes to AI infrastructure, stock surges 580% in single day

Allbirds, once known for sustainable footwear, announced a dramatic shift to become NewBird AI, a GPU-as-a-Service provider. After selling its shoe brand for $39 million to American Exchange Group, the company secured $50 million in financing to acquire GPU assets. The announcement sent shares soaring over 580%, though questions remain about the execution of this extreme pivot.

47 Sources

News websites block Wayback Machine as AI scraping fears threaten digital preservation

Major news outlets including the New York Times and USA Today are blocking the Wayback Machine from archiving their content, citing fears that AI companies exploit the archive to train models. The move threatens the Internet Archive's ability to preserve the public record, even as some of these same publishers rely on the archive for their own investigative reporting.

2 Sources

Starbucks launches ChatGPT app to suggest drinks based on mood, outfit, and customer vibe

Starbucks rolled out a beta app inside ChatGPT on Wednesday, enabling customers to discover personalized drink recommendations based on feelings, photos, or current mood. While orders start in the AI chatbot, customers must complete purchases through the Starbucks app or website. The move signals how major brands are embedding AI into customer discovery.

14 Sources

Artemis Raises $70M to Fight AI-Powered Attacks With AI-Native Security Platform

Six-month-old cybersecurity startup Artemis emerged from stealth with $70 million in combined seed and Series A funding to combat AI-powered attacks that move at machine speed. Founded by former AWS and Abnormal Security leaders, the company offers an AI-native protection platform that replaces traditional rule-based systems with adaptive, automated threat response.

4 Sources

EU threatens interim ban on Meta over WhatsApp AI policies blocking rival AI chatbots

The European Commission rejected Meta's pay-for-access remedy for third-party AI providers on WhatsApp, threatening interim measures to restore full access. The EU antitrust watchdog says Meta's fee-based approach is equivalent to an outright ban and constitutes abuse of dominant position, potentially harming competition in the AI assistant market.

7 Sources

Pharos Network launches AI research project with Hong Kong to explore prediction markets

Pharos Network has partnered with the University of Hong Kong's FinTech Academy on a three-month research initiative exploring AI decision-making in prediction markets. Eight Master's students will analyze real onchain data to test whether AI models can predict event probabilities more accurately than humans, with promising projects fast-tracked into Pharos's incubation program for live deployment.

2 Sources

Jane Street commits $6 billion to CoreWeave AI cloud, takes $1 billion equity stake

Quantitative trading firm Jane Street has signed a $6 billion AI cloud agreement with CoreWeave and invested $1 billion in equity at $109 per share. The deal grants Jane Street access to Nvidia's next-generation Vera Rubin chips and marks CoreWeave's third multi-billion-dollar agreement this month, following deals with Meta and Anthropic.

8 Sources

Trump backs AI safeguards in banking system, acknowledges risks after cybersecurity warnings

President Donald Trump acknowledged that artificial intelligence could undermine confidence in the banking system and endorsed government safeguards, including a potential kill switch. His comments follow cybersecurity warnings about Anthropic's Mythos AI model potentially supercharging complex cyberattacks against banking infrastructure.

2 Sources

Emergent's Wingman AI agent tackles routine tasks through WhatsApp and Telegram messaging

SoftBank-backed Emergent has launched Wingman, an autonomous AI agent that automates routine tasks across popular messaging platforms like WhatsApp and Telegram. Unlike competitors, Wingman operates with trust boundaries, seeking user confirmation for consequential actions while handling low-stakes tasks independently. The move signals Emergent's expansion beyond its vibe-coding platform into the rapidly growing AI agent space.

3 Sources

Duolingo reverses course on AI performance reviews after employees push back on metrics

Duolingo CEO Luis von Ahn has abandoned plans to evaluate employees based on how much they use AI at work, admitting the policy created confusion about whether tool usage mattered more than actual results. The reversal comes after significant internal pushback and marks the second time in a year the company has backtracked on its AI-first strategy.

3 Sources

AI agents from Anthropic, Google, and Microsoft hijacked via prompt injection, no CVEs issued

Security researchers hijacked AI agents from Anthropic, Google, and Microsoft through prompt injection attacks on GitHub Actions integrations, stealing API keys and tokens. All three companies paid bug bounties—$100 from Anthropic, $500 from GitHub—but none published advisories or assigned CVEs, leaving users on older versions unaware of the risk.

3 Sources

Tesla completes AI5 chip design with 40X performance leap, but deployment still years away

Elon Musk announced Tesla's successful tape-out of the AI5 chip, designed to power Full Self-Driving systems in vehicles and Optimus robots. The chip promises up to 40x performance improvement over its predecessor, featuring 12 memory packages and a 384-bit interface. However, the milestone comes nearly two years behind schedule, with volume production not expected until mid-2027.

6 Sources

North Korean Hackers Use AI Social Engineering in $100,000 Crypto Attack on Zerion

North Korean-affiliated hackers used AI social engineering to steal $100,000 from crypto wallet company Zerion's hot wallets. The attack marks the second major DPRK-linked breach this month, following Drift Protocol's $280 million exploit. Security experts warn that AI-driven threats are making attacks more sophisticated and harder to detect.

2 Sources

Get all updates into your inbox 💌, Stay Up-to-date for Free 🥳

Your Daily Dose of Curated AI News

Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Spend less time searching for the latest in AI and get straight to action.

The Outpost

News

About