anthropic launches openai while google expands its platform

Key Takeaways

• Anthropic is limiting public access to its new AI model, Mythos (Claude Mythos), due to its advanced ability to autonomously find and exploit zero-day cybersecurity vulnerabilities.
• Mythos is being shared with select companies through Project Glasswing to help strengthen defenses and prepare for advanced AI-powered cyberattacks.
• Bank CEOs, the Federal Reserve Chair, and the Treasury Secretary met with Anthropic to discuss Mythos's potential to destabilize the financial system.
• OpenAI is reportedly considering similar restrictions for its upcoming AI models due to cybersecurity concerns.
• Google's AI Mode in Search now offers agentic capabilities in the UK, simplifying restaurant bookings by searching multiple platforms based on user needs.
• OpenAI added voice interaction features to ChatGPT, including Voice Mode for real-time conversation and Dictation mode for hands-free use.
• AutoRaptor launched an AI Voice Agent for car dealerships, providing 24/7 automated lead capture, inquiry handling, and appointment scheduling.
• The UK government is expanding its AI Accelerator program, which has trained 49 civil servants since March 2025, to build in-house AI capabilities.
• AI tools that identify speakers by voice may violate biometric privacy laws like Illinois' BIPA by collecting voiceprints without explicit consent.
• A webinar on April 15, 2026, will discuss enterprise AI security in the "agentic era," focusing on securing autonomous agents and the AI supply chain.

Anthropic limits new AI model Mythos due to cybersecurity risks

Anthropic is restricting access to its new AI model, Mythos, because it poses significant cybersecurity risks. The company wants to give cyber defenders more time to strengthen their systems. However, many experts believe AI's cyber risks have already reached a dangerous point, even with current models. These advanced AI systems can automate complex cyberattacks, making them accessible to people with less technical skill. This raises concerns about protecting critical infrastructure from potential attacks on systems like banks and power grids.

New AI model Mythos raises stakes for cybersecurity experts

Anthropic's new AI model, Claude Mythos, is signaling an AI arms race in cybersecurity. The company is limiting its release to industry partners to prevent misuse, a move also reportedly being considered by OpenAI for its new model. Experts note that advanced AI can now find vulnerabilities at a speed and scale previously only possible for top security professionals. While this benefits defenders, it also means new AI-powered threats and faster attacks are emerging, requiring security teams to adapt to a rapidly evolving landscape.

Anthropic's powerful Mythos AI is a cybersecurity wake-up call

Anthropic is withholding its new AI model, Mythos, from public release due to its advanced ability to find software vulnerabilities. The company is sharing the technology with select major companies through Project Glasswing to help them improve their defenses. Security experts warn that AI's potential for harm is already here, and organizations must prepare for faster and more capable cyberattacks. Government officials, including the Treasury Secretary and Federal Reserve Chair, have been briefed on these risks, highlighting the severe potential consequences for economies and national security.

Mythos AI forces cybersecurity reckoning for developers

Anthropic's new AI model, Mythos Preview, can discover software vulnerabilities and develop exploits autonomously. Due to its power, Anthropic is only releasing it to a select group of organizations through Project Glasswing. While some are skeptical, many experts believe this AI capability, especially in creating exploit chains, is a real threat. This limited release gives defenders a head start to prepare for a future where attackers may have widespread access to such powerful tools, pushing developers to prioritize security.

Mythos AI finds major security flaws in operating systems and browsers

Anthropic's new AI model, Mythos, can identify and exploit zero-day vulnerabilities in major operating systems and web browsers, according to the company. This capability, which previously required expert professionals, can now be achieved in hours by engineers without security backgrounds. Anthropic is not releasing Mythos publicly due to its potent hacking abilities and is instead sharing it with select companies through Project Glasswing. The model's ability to chain vulnerabilities and operate autonomously marks a significant shift in cybersecurity, potentially weakening defenses that rely on friction.

Bank CEOs meet over Anthropic's Mythos AI security risks

Bank CEOs met with Anthropic executives, the Federal Reserve Chair, and the Treasury Secretary to discuss the security risks posed by Anthropic's new AI model, Claude Mythos. This model can find software flaws, including old vulnerabilities, much faster than humans. The concern is that bad actors could use Mythos to exploit coding weaknesses before banks can fix them, potentially destabilizing the financial system. Anthropic has not widely released Mythos, instead forming Project Glasswing with major tech companies to find and fix bugs before they can be exploited.

Webinar to cover AI security in the agentic era of 2026

A webinar on April 15, 2026, will discuss the evolving landscape of AI security for enterprises. It will cover the shift from AI experimentation in 2025 to autonomous, enterprise-wide deployments in 2026, focusing on agentic AI where models execute actions. Key topics include securing autonomous agents, protecting the AI supply chain, navigating global AI regulations, and using predictive defense. The event aims to provide practical guidance for IT and security leaders to build secure AI infrastructure and manage risks.

Webinar details AI security in the agentic era

A webinar scheduled for April 15, 2026, will explore enterprise AI security in the agentic era. The session aims to clarify the real risks, responsibilities, and responses shaping AI adoption. It will cover topics such as securing autonomous agents, protecting the AI supply chain, and adapting to new global regulations. The webinar targets leaders and practitioners needing strategic and operational guidance for managing AI systems and security.

AutoRaptor launches AI Voice Agent for car dealerships

AutoRaptor has launched an AI Voice Agent to handle inbound calls for automotive dealerships 24/7. This automated system captures leads, answers questions about inventory and financing, and schedules appointments directly within the AutoRaptor CRM. The goal is to prevent lost sales from unanswered calls, especially during busy times or after hours. The AI agent can also filter out spam and non-sales calls, with future updates planned for proactive follow-ups and deeper integration with AutoRaptor's AI Sales Assistant.

UK government expands AI training for public sector skills

The UK government is expanding its AI Accelerator program to build in-house data and machine learning capabilities within government departments. Since its launch in March 2025, the program has trained 49 civil servants in AI and data skills. The initiative focuses on hands-on learning, with participants developing AI solutions like chatbots and large language models. Now, departments can fund their own tailored versions of the program, reflecting a shift towards developing owned AI capabilities to meet specific operational needs.

UK restaurant bookings now easier with Google AI Mode

Google has introduced new agentic capabilities to AI Mode in Search for users in the UK, making restaurant bookings simpler. Users can now describe their needs, such as group size, dietary restrictions, and location, and AI Mode will search multiple booking platforms for real-time availability. The feature presents curated options with direct links to partners like TheFork and Sevenrooms. This update aims to reduce the time spent planning and streamline the process from searching to booking.

AI meeting tools may violate biometric privacy laws

AI tools that identify speakers by analyzing their voices may be collecting biometric data without proper consent, potentially violating laws like Illinois' Biometric Information Privacy Act (BIPA). A recent lawsuit highlighted how these tools might create and store voiceprints without notice or written consent, even from non-users. BIPA requires written notification, disclosure of purpose and duration, and a written release before collecting biometric identifiers. Recent court rulings suggest liability may depend on whether an entity actually controls the biometric data.

AI is transforming retail experiences in Mexico

In Mexico's competitive retail market, AI is becoming crucial for differentiating through personalized shopping experiences. A study found that many leading retailers struggle with key friction points, especially when customers show purchase intent. Issues include internal search engines lacking semantic understanding and error tolerance, and chatbots with limited functionality. The study highlights the need for agentic commerce, where AI understands context and various information types for a seamless customer journey, as many current chatbots fail to transfer conversations to humans without losing context.

Man uses AI to help treat girlfriend's brain tumor

After standard medical treatments failed to cure his girlfriend Amy's recurring brain tumor, Andrew Rodriguez turned to AI for answers. Amy was diagnosed with a prolactinoma, a pituitary gland tumor, which had returned despite two surgeries. Frustrated with the medical system, Rodriguez began using an AI chatbot, leading him to believe he could find a way to cure her himself. The article details his journey and the unconventional approach he took to combat the illness.

Bild AI seeks founding product engineer for construction workflow automation

Bild AI, a startup focused on automating tasks in the construction industry like blueprint reading, cost estimation, and permit applications, is hiring a founding product engineer. The company aims to streamline the entire process from blueprint to permit through end-to-end automation and product-led engineering. Bild AI is looking for someone to help build its automated solutions for the construction workflow.

ChatGPT now offers voice interaction for hands-free use

OpenAI has added voice features to ChatGPT, allowing users to interact with the AI through speech. This enables hands-free operation, making conversations faster and more natural. Users can choose between Voice Mode for real-time conversation or Dictation mode to convert speech to text. The new features aim to improve efficiency during multitasking and commuting, offering a more accessible way to use ChatGPT. Audio and video clips from voice chats are saved in the chat history.

Sources

• Anthropic is limiting access to its latest AI model, Mythos. The real risks may already be out there
• AI is raising the stakes for cyber professionals – Claude Mythos just took things to another level
• Anthropic's potent new AI model is a "wake-up call," security experts say
• Anthropic’s Mythos Will Force a Cybersecurity Reckoning—Just Not the One You Think
• Anthropic’s Mythos AI Uncovered Serious Security Holes in Every Major OS and Browser
• Why Anthropic met with bank CEOs about AI security risks
• Unpacking AI security in 2026 from experimentation to the agentic era
• Webinar Explains AI Security in Agentic Era
• AutoRaptor Announced the Launch of AI Voice Agent
• UK government expands AI Accelerator program for workforce skills | ETIH EdTech News
• Booking restaurants in the UK just got easier with AI in Search
• AI Meeting Tools Are Testing Biometric Privacy
• AI in Retail: Transforming the Shopping Experience
• I Used AI to Fight My Girlfriend’s Brain Tumor
• Bild AI Seeks Founding Product Engineer
• ChatGPT Adds Voice Interaction