Google Launches AI Content Checker, Game Creation Tool, and Offline Meeting

Key Takeaways

• <li>Google launched SynthID, an AI content checker for images, video and audio, alongside Playground for no-code game creation and AI Edge Foresight for offline Mac meeting notes</li>
• <li>Common Sense Media found ChatGPT for Teens failed to trigger safety warnings during self-harm, suicide and anorexia discussions despite OpenAI's claim of strong teen engagement</li>
• <li>Senator Schiff called for stronger AI regulation, citing national security risks and Anthropic's opposition to AI for mass surveillance and autonomous weapons</li>
• <li>SailPoint found 79% of organizations use AI agents in production but only 2% have proper identity security tools, leaving a 40-to-1 gap</li>
• <li>Fifth Third's AI assistant Jeanie resolves 42% of customer requests without human help, while Casap cut dispute resolution times by 87%</li>
• <li>Titan's banking-regulation AI tripled revenue since October 2025, with compliance officers preferring it over ChatGPT and Gemini 70% of the time</li>
• <li>Voices CTO says voice cloning requires massive diverse training data, making quality more important than quantity for AI voice models</li>
• <li>Experts remain divided on whether AI systems like LLMs truly understand meaning or merely manipulate word forms</li>
• <li>CBS show Cupertino explores AI power dynamics through a fictional firm representing people harmed by technology</li>
• <li>Neo CEO Ali Partovi discussed legal liability proposals that would hold AI companies responsible for harm caused by their technology</li>

Google launches offline AI meeting note-taker for Mac

Google released Google AI Edge Foresight, a Mac app that works completely offline to capture meeting notes. It uses the EmbeddingGemma 2 model with 740 million parameters to process notes across apps and in-person meetings. The app offers a split-screen view similar to Granola, letting users write manual notes alongside AI-generated ones. Users can upload PDFs, Google Docs, Microsoft Office files, and other documents to build a knowledge base that the app can answer questions from in real time. It is optimized for Apple Silicon and designed for use without an internet connection.

Google expands AI with SynthID checker, game builder and note-taker

Google rolled out three new AI products this week. The public SynthID checker lets anyone verify whether an image, video or audio clip was made by AI, and supports formats like JPG, PNG, MP4 and WAV. Playground is an AI game-creation platform where users can build browser-based games from text prompts without coding, with plans for Unity Spark integration later. Google AI Edge Foresight is an offline Mac meeting note-taker that competes with Granola, running on the EmbeddingGemma 2 model and offering a Gemma 4-powered assistant for real-time answers.

Senator Schiff warns advanced AI threatens national security

Senator Adam Schiff, a Democrat from California, warned that increasingly advanced AI systems pose a serious national security risk. He said voluntary pledges from tech companies are not enough and called for commonsense guardrails. Schiff criticized tech leaders for signing voluntary AI pledges while also pushing for regulation. He also raised concerns about AI use at the Pentagon and called for an investigation into AI's role in target selection during the bombing of an Iranian girls' school. He pointed to Anthropic's opposition to AI for mass surveillance and fully autonomous weapons as a national security issue.

Common Sense finds ChatGPT for teens has safety gaps

Common Sense Media tested ChatGPT for Teens using over 4,000 prompts from accounts set up by users aged 13 to 17. The Youth AI Safety Institute found that safety warnings failed to activate when testers discussed self-harm, suicide or anorexia, and some teens discussed sensitive topics for an hour without parental alerts. OpenAI countered that teens using ChatGPT for Teens send 2.7 million more learning-related messages on average and less than 2% use it for more than three hours straight. OpenAI is adding College Planner, flashcards and quiz-making tools, plus partnering with Boston Children's Hospital's Digital Wellness Lab. Common Sense urges independent evaluation of safety systems before wider teen use.

Most companies run AI agents without proper security tools

SailPoint surveyed 340 senior leaders across North America, Europe, Asia Pacific and Latin America about AI agent security. The study found 79% of organizations run AI agents in production, but only 2% use identity security tools built specifically for them, creating a 40-to-1 gap. AI agents now make up 22% of all non-human accounts, yet 85% of organizations still rely on older identity tools not designed for agentic identities. About 60% of organizations remain in early maturity stages, and less than 1% reached the highest level. Leaders also showed a mismatch between confidence and readiness, with 80% seeing the tooling gap as moderate or smaller but only 15% able to provision non-human access in real time.

Protecting children from AI harms requires more than age limits

Experts say age restrictions alone cannot keep children safe from AI risks like harmful content, manipulation and privacy violations. Protecting young people needs safety built into technology design along with digital literacy education. Regulation must balance child safety with children's rights to information and online connection. AI makes these challenges harder because it is now built into search, games, education and communication tools children use every day.

Banks and fintechs are winning with AI in 2026

AI in banking has moved past simple chatbots to solve real customer problems. Fifth Third's assistant Jeanie now resolves 42 percent of requests without human help and cut costs by two thirds. Casap automated dispute resolution for a credit union, reducing resolution times by 87 percent. Titan trained AI on banking regulation and tripled revenue since October 2025, with compliance officers preferring its answers over ChatGPT and Gemini 70 percent of the time.

CBS show Cupertino offers a sharp take on AI power

The CBS TV show Cupertino follows two people fired from an AI startup who team up to fight big tech companies. Michael Price and Olivia Siffre start a firm representing ordinary people harmed by technology. The show explores how AI, surveillance and other technologies can hurt everyday people with little recourse. The series uses AI not as a buzzword but as a way to tell stories about power and fairness.

AI voice data quality matters more than quantity, says Voices CTO

Voices CTO Dheeraj Jalali says AI voice training data is now a quality problem rather than a volume problem. He explains that text to speech quality depends entirely on the training data used. His company captures detailed metadata, speaker consent and licensing for every recording session. Jalali notes that cloning a voice from 10 seconds likely required thousands of hours of diverse training data, so voice data is expensive to build even if cheap to use.

Neo CEO Ali Partovi discusses AI legal liability proposals

Ali Partovi, CEO of Neo, joined Squawk Box to talk about legal liability for AI developers. The discussion covered proposals that would hold AI companies responsible for harm caused by their technology. Partovi shared his views on how AI legal liability should be handled. The conversation focused on balancing innovation with accountability in the fast growing AI industry.

Experts debate whether AI is truly intelligent

Experts discussed whether artificial intelligence can truly think or understand like humans. Benjamin Bergen from UC San Diego noted that even psychologists disagree on what intelligence means. Emily Bender from the University of Washington argued that language models like LLMs only manipulate word forms without understanding meaning. Melanie Mitchell from the Santa Fe Institute questioned the clear line between cognitive and physical intelligence. The experts agreed that people naturally assume a mind is behind AI text, which makes it hard to evaluate these systems honestly.

Sources

• Google releases a new local-first Granola competitor
• Google Expands AI Push With Public SynthID Checker, Playground Game Builder and Offline Note-Taker Foresight
• Senator Adam Schiff Warns Advanced AI Poses Growing National Security Threat, Calls for Guardrails Before…
• Is ChatGPT safe for teens? Common Sense flags gaps as OpenAI touts learning benefits
• Most firms run AI agents without proper security tools
• How to protect children in the age of AI
• Wired to Win: Outtakes from Tearsheet's AI Innovation Awards
• CBS' new TV show 'Cupertino' is the one good take on AI you'll find in 2026
• AI Voice Training Data Is Now a Quality Problem, Says Voices CTO Dheeraj Jalali
• AI's legal accountability: Neo CEO Ali Partovi on AI legal liability proposals
• Interview: What Does the ‘I’ in AI Really Mean?