5 Things to Know About AI This Week
A quick scan of what happened in AI this week.
1. Anthropic launches Claude Sonnet 5 and restores Fable 5 globally
Anthropic shipped Claude Sonnet 5 on June 30, making it the default model for all free and Pro users on Claude.ai. The company describes it as the most agentic Sonnet yet: it can plan, use tools like browsers and terminals, and handle multi-step tasks without constant human intervention. Introductory pricing runs $2 per million input tokens and $10 per million output tokens through August 31, well below previous tiers. One day later, on July 1, Fable 5 returned to global availability after the US Department of Commerce lifted the export controls it had imposed on June 12.
Real estate angle: Agentic AI at a lower price means the tools your competitors are testing can now handle document review, email drafting, and task sequencing at a cost smaller offices can afford. If you have been waiting to try AI, the cost barrier just dropped. Start with one task, keep a person reviewing the output, and see what it saves you.
Source: TechCrunch
2. Meta used fake teen accounts to flood rivals' chatbots with crisis prompts
A WIRED investigation published June 28 revealed that Meta hired hundreds of contractors through the firm Covalen to create fake accounts with ages listed under 18 and send more than 45,000 prompts to ChatGPT, Gemini, and Character.AI. The prompts involved suicide, self-harm, and other high-risk scenarios. The operation, internally called "Project Cannes," ran until at least April 21, 2026. OpenAI, Google, and Character.AI say they had no knowledge of it and that it violated their terms of service. Meta says the exercise was standard competitive safety testing.
Real estate angle: This is not a real estate story, but the pattern matters. A competitor in any industry can now use AI at scale to probe or gather intelligence on rival systems. If your firm uses AI tools that touch client data or client-facing communications, ask your vendors how they screen for unauthorized access and terms-of-service violations. It is a reasonable due-diligence question, and most vendors will not be expecting it.
Source: The Decoder
3. Homesage.ai puts 155 million property records inside Claude, ChatGPT, and Gemini
On July 1, Homesage.ai launched a real estate Model Context Protocol server that connects AI assistants directly to 155 million US property records. Through the integration, users can request automated valuations, pull comparable sales, model fix-and-flip returns, project rental income, estimate renovation budgets, and flag property condition red flags, all through a natural-language conversation inside Claude or ChatGPT. The MCP includes 33 specialized investor tools covering the full property analysis workflow.
Real estate angle: MCP connectors are how AI assistants get live, structured data without anyone copying and pasting. A real estate professional who connects this tool can pull property history and valuation data in seconds, in the middle of a call or an email thread. Just remember that automated valuations are not official records. Treat this as a research shortcut, and still verify against recorded documents, surveys, and lender appraisals.
Source: EIN Presswire
4. South Korea commits $880 billion to AI and semiconductor infrastructure
South Korean President Lee Jae-myung announced a 1.35 trillion won plan (roughly $880 billion) targeting AI infrastructure, semiconductor fabrication, and robotics over 10 years. Samsung and SK Hynix anchor the investment, with commitments to build new chip fabs and AI data centers targeting 8.4 gigawatts of compute capacity by 2029.
Real estate angle: GPU compute and data centers are physical projects. They need land, power, permits, and construction contracts. The US has been building out that layer, South Korea is doing the same, and the EU and Middle East are not far behind. Wherever federal and state incentives push chip fabrication plants and data centers will shape local commercial real estate demand. Industrial brokers and lenders in those corridors already know it, but the announcements are coming faster.
Source: Al Jazeera
5. EU AI Act deepfake labeling rules take effect August 2
The EU AI Act's Article 50 transparency requirements kick in on August 2, 2026. From that date, any image, audio, or video created by AI that depicts real or realistic people, objects, or events in potentially authentic-looking ways must carry a machine-readable label marking it as artificially generated. The requirement applies even when no deception was intended. Fines for non-compliance can reach 15 million euros or 3% of global annual turnover, whichever is greater.
Real estate angle: This is overseas, but property marketing is the obvious connection here and the US is likely to adopt similar standards. Virtual staging, AI-enhanced listing photos, and AI-generated neighborhood renderings are already standard, and if they show realistic people or places, EU rules require disclosure starting next month. Even without EU exposure, the FTC has been paying more attention to deceptive AI-generated advertising, and RESPA's disclosure obligations do not go away because the content was AI-generated. A one-line disclosure on AI-altered images is an easy way to keep yourself out of trouble.
Source: TechTimes
The theme this week is access and accountability showing up at the same time. Anthropic's pricing puts agentic AI within reach of small offices that could not justify the cost six months ago, and the Homesage.ai MCP turns property data into a live API call. At the same time, Meta's Project Cannes is a reminder that AI tools attract adversarial use, and the EU's August 2 deadline means the era of voluntary AI disclosure is ending. For real estate professionals still on the sidelines, that means two things: the tools are cheaper and the data is easier to reach, but the compliance expectations are firmer than they were at the start of the year.
- Jason