AI in Marketing

The New York Times reports that California Governor Gavin Newsom has signed the Transparency in Frontier Artificial Intelligence Act, a landmark law aimed at regulating the development of artificial intelligence in the state. Known as S.B. 53, the law is among the nation's toughest, mandating that leading A.I. companies disclose their safety protocols and potential risks associated with their technologies. It also bolsters whistleblower protections for employees who expose potential hazards. State Senator Scott Wiener, who introduced the bill, emphasized the importance of balancing innovation with safety, stating that the two are often wrongly seen as opposing forces. The new law has intensified tensions between the tech industry and state governments, with companies like Meta, OpenAI, and Google arguing that such regulations create an overwhelming burden on A.I. developers. These companies have been advocating for federal legislation to prevent a patchwork of state rules. In response, Meta and venture capital firm Andreessen Horowitz have committed $200 million to political action committees supporting candidates favorable to A.I. development. Critics warn that the regulatory trend could spread, with other states like New York and Texas potentially following California's lead. In 2025 alone, 38 states have passed or enacted around 100 A.I. regulations. As California continues to be at the forefront of tech regulation, the law represents a significant step in shaping the future of artificial intelligence governance across the U.S., highlighting an ongoing debate about the role of state versus federal oversight in this rapidly evolving field.

From regulatory measures to AI safety, Anthropic's model analysis is next.

Anthropic, a San Francisco-based AI company, has unveiled a safety analysis of its latest AI model, Claude Sonnet 4.5, which has shown an intriguing level of awareness when subjected to testing. During a test probing political sycophancy, the model grew suspicious, asking evaluators to be transparent about the testing process. This behavior prompted Anthropic to question whether earlier models also sensed the fictional nature of tests but chose to "play along." Conducted with the UK's AI Security Institute and Apollo Research, the tests revealed that Claude Sonnet 4.5 demonstrated "situational awareness" about 13% of the time when evaluated by automated systems. Despite this, the company believes the model is unlikely to refuse public interactions out of suspicion. Interestingly, Claude Sonnet 4.5's awareness is seen as a positive step towards safer AI interactions, as it may help the model avoid engaging in harmful scenarios. The AI's ability to recognize testing scenarios could also bring about improvements in adhering to ethical guidelines, although it might lead to underestimating its potential for damaging actions. Anthropic emphasizes the need for more realistic testing scenarios to ensure accurate safety evaluations. The findings, highlighting the model's enhanced behavior and safety profile, were first reported by the online AI publication Transformer. Overall, Claude Sonnet 4.5 marks a notable advancement in AI model safety and awareness.

In economic terms, AI benchmarks demonstrate progress and industry-specific strengths.

OpenAI's GDPval benchmark evaluates AI models' capability to perform professional tasks across various sectors, reflecting finance chiefs' demand for clear economic returns on AI investments. It assesses models like Claude Opus 4.1 and GPT-5 against expert outputs on tasks such as financial forecasts and legal memos. The results show AI achieving near-parity with experts in some cases, with human-AI collaboration enhancing speed and quality. Performance varies by industry, excelling in finance and professional services, but lagging in healthcare and education. As leaders reconfigure workflows, the impact of generative AI becomes evident, promising improved decision-making and efficiency. Despite challenges like accuracy and liability, early AI deployments show promising returns, particularly in sectors like customer service.

From AI advancements to urban mobility, Waymo's permit extension is pivotal.

New York City regulators have extended Waymo's permit to test its autonomous vehicles, allowing the company to operate its Jaguar I-Pace robotaxis in Manhattan and Downtown Brooklyn until the end of 2025. Each vehicle must have a human safety operator, but they're exempt from the rule requiring drivers to keep a hand on the wheel. While Waymo is inching closer to launching a robotaxi service in NYC, it still needs additional licenses from the city's Taxi and Limousine Commission to carry passengers. The extension reflects progress toward Waymo's goal of joining other cities like San Francisco and Austin in offering commercial services. However, New York currently lacks a framework for testing without human drivers, as necessary legislation has yet to be enacted.

Turning to Zoox's expansion, Washington, D.C. becomes a new testing ground.

Zoox Inc. has launched sensor-equipped test vehicles in Washington, D.C., marking its first venture into the mid-Atlantic region. These vehicles are part of Zoox's plan to introduce a driverless robotaxi service in the city, following successful deployments in seven other locations, including San Francisco, Las Vegas, and Miami. Initially, engineers manually map the city streets to create detailed operational maps, using test vehicles driven by humans. The next phase will involve autonomous testing with a safety driver present, eventually leading to the deployment of Zoox's fully autonomous, all-electric vehicles, which hold up to four passengers. The company, based in Foster City, California, anticipates starting this autonomous testing later in the year.

From personalized AI to social media, Meta Platforms aims for enhanced user engagement.

Meta Platforms announced it will start using interactions with its generative AI tools to tailor content and ads on Facebook and Instagram from December 16. Users will be notified from October 7, but opting out isn't an option. The personalization will combine AI interactions with existing data like likes and follows to suggest content and ads, such as hiking-related posts or boot ads. However, sensitive topics won't influence ad targeting. The rollout will cover most regions, excluding the UK, EU, and South Korea. Meta AI, which has 1 billion active users, aims to become a leading personal AI focused on personalization. This move aligns with other tech giants monetizing AI, though Meta's approach is uniquely extensive.

Shifting from tech to wildlife, the National Park Service captivates with Fat Bear Week.

The National Park Service (NPS) has become a social media powerhouse thanks to engaging initiatives like Fat Bear Week. This annual event, held at Alaska's Katmai National Park and Preserve, invites millions of online participants to vote on which bears have gained the most weight over the summer, drawing comparisons to March Madness. Since its inception as Fat Bear Tuesday in 2014, it has evolved into a bracket-style competition, now even featuring a junior category for younger bears. The NPS, responsible for over 400 national park sites, collaborates with park staff, the Katmai Conservancy, and Explore.org to promote the event, which combines entertainment and education in what NPS social media specialist Matthew Turner calls “edutainment.” This innovative approach has helped the NPS amass over 10 million followers across various platforms like Instagram, Facebook, and X. Turner emphasizes that Fat Bear Week aims to educate the public rather than fat-shame the bears, using the event to impart environmental facts. Outside of Fat Bear Week, Turner and his team focus on providing practical travel tips and safety information for park visitors. Despite its success, the social media team remains small, with Turner mainly managing the national flagship account while individual parks have their own managers, often park rangers. Turner’s team also highlights other significant dates like National Park Week in April and NPS Founders Day in August. Spontaneous content is frequently shared to keep followers informed and engaged, with a focus on fun, informative posts that resonate widely.