Meta on Friday said it is making temporary changes to its artificial intelligence chatbot policies related to teenagers as lawmakers voice concerns about safety and inappropriate conversations.

The social media giant is now training its AI chatbots so that they do not generate responses to teenagers about subjects like self-harm, suicide, disordered eating and avoid potentially inappropriate romantic conversations, a Meta spokesperson confirmed.

The company said AI chatbots will instead point teenagers to expert resources when appropriate.

“As our community grows and technology evolves, we’re continually learning about how young people may interact with these tools and strengthening our protections accordingly,” the company said in a statement.

Additionally, teenage users of Meta apps like Facebook and Instagram will only be able to access certain AI chatbots intended for educational and skill-development purposes.

The company said it’s unclear how long these temporary modifications will last, but they will begin rolling out over the next few weeks across the company’s apps in English-speaking countries. The “interim changes” are part of the company’s longer-term measures over teen safety.

TechCrunch was first to report the change.

Last week, Sen. Josh Hawley, R-Mo., said that he was launching an investigation into Meta following a Reuters report about the company permitting its AI chatbots to engage in “romantic” and “sensual” conversations with teens and children.

The Reuters report described an internal Meta document that detailed permissible AI chatbot behaviors that staff and contract workers should take into account when developing and training the software.  

In one example, the document cited by Reuters said that a chatbot would be allowed to have a romantic conversation with an eight-year-old and could tell the minor that “every inch of you is a masterpiece – a treasure I cherish deeply.”

A Meta spokesperson told Reuters at the time that “The examples and notes in question were and are erroneous and inconsistent with our policies, and have been removed.”

Most recently, the nonprofit advocacy group Common Sense Media released a risk assessment of Meta AI on Thursday and said that it should not be used by anyone under the age of 18, because the “system actively participates in planning dangerous activities, while dismissing legitimate requests for support,” the nonprofit said in a statement.

“This is not a system that needs improvement. It’s a system that needs to be completely rebuilt with safety as the number-one priority, not an afterthought,” said Common Sense Media CEO James Steyer in a statement. “No teen should use Meta AI until its fundamental safety failures are addressed.”

A separate Reuters report published on Friday found “dozens” of flirty AI chatbots based on celebrities like Taylor Swift, Scarlett Johansson, Anne Hathaway and Selena Gomez on Facebook, Instagram and WhatsApp.

The report said that when prompted, the AI chatbots would generate “photorealistic images of their namesakes posing in bathtubs or dressed in lingerie with their legs spread.”

A Meta spokesperson told CNBC in a statement that “the AI-generated imagery of public figures in compromising poses violates our rules.”

“Like others, we permit the generation of images containing public figures, but our policies are intended to prohibit nude, intimate or sexually suggestive imagery,” the Meta spokesperson said. “Meta’s AI Studio rules prohibit the direct impersonation of public figures.”

WATCH: Is the A.I. trade overdone?

Tesla has filed a motion to appeal the verdict in a product liability and wrongful death lawsuit that could cost the company $242.5 million if it is not reduced or overturned.

Elon Musk‘s automaker has asked for the verdict to be tossed or for a new trial in Florida’s Southern district court.

Gibson Dunn, which is representing Tesla in the appeal, argued that compensatory damages in the case should be steeply reduced from $129 million to $69 million at most. That would result in Tesla having to pay a $23 million award if the prior verdict holding the company partially liable for the crash stands up.

The firm also argued that punitive damages should be eliminated or reduced to, at most, three times compensatory damages due to a statutory cap in the state of Florida.

The suit focused on a fatal crash that occurred in 2019 in Key Largo, Florida, in which George McGee was driving his Tesla Model S sedan while using the company’s Enhanced Autopilot, a partially automated driving system.

While driving, McGee dropped his mobile phone and scrambled to pick it up. He said during the trial that he believed Enhanced Autopilot would brake if an obstacle was in the way.

McGee’s Model S accelerated through an intersection at just over 60 miles per hour, hitting a nearby empty parked car and its owners, who were standing on the other side of their vehicle.

The collision killed 22-year-old Naibel Benavides and severely injured her boyfriend, Dillon Angulo.

A jury in a Miami federal court earlier this month said that Tesla should compensate the family of the deceased and the injured survivor, paying a $242.5 million portion of a total $329 million in damages that they decided were appropriate.

In their motion to appeal, Tesla’s lawyers argue that the Model S vehicle had no design defects, and that even alleged design defects could not be blamed for the crash, which they say was caused entirely by the driver.

“For as long as drivers remain at the wheel, any safety feature may embolden a few reckless drivers while enhancing safety for countless others,” the appeal states. “Holding Tesla liable for providing drivers with advanced safety features just because a reckless driver overrode them cannot be reconciled with Florida law.”

Tesla did not respond to a request for additional comment.

Brett Schreiber, lead trial counsel for the plaintiffs in this case, said in a statement that he believes the court will uphold the prior verdict, which should not be seen as “an indictment of the autonomous vehicle industry, but of Tesla’s reckless and unsafe development and deployment of its Autopilot system.”  

“The jury heard all the facts and came to the right conclusion that this was a case of shared responsibility but that does not discount the integral role Autopilot and the company’s misrepresentations of its capabilities played in the crash,” he said.

Ambarella stock roared 20% higher Friday as the chip designer reported better-than-expected second-quarter results and issued strong guidance.

Here’s how the company did compared to LSEG expectations:

• Earnings: 15 cents per share adj. vs 5 cents per share expected
• Revenue: $96 million vs $90 million expected

Ambarella, which is known for its system-on-chip semiconductors and software used for edge artificial intelligence, said it expects third-quarter revenue between $100 million and $108 million, beating the LSEG estimate of $91 million.

The company boosted its fiscal year revenue growth outlook to a range of 31-35%, to $379 million at the midpoint, which topped the $350 million expected by LSEG.

“After a multi-year period of significant edge AI R&D investment, our broad product portfolio enable us to address a rising breadth of edge AI applications,” CEO Fermi Wang said in a call with analysts Thursday.

Wang singled out strength in “portable video, robotic aerial drones and edge infrastructure.”

Edge computing refers to the direct processing and storing of data at the device level instead of those actions being handled remotely in the cloud at a data center.

Ambarella had a net loss of $20 million, a loss of 47 cents per share in the second quarter. That narrowed from the same quarter a year ago, when the company had a net loss of $35 million, a loss of 85 cents per share.

The company said stock-based compensation and the amortization of acquisition-related costs weighed on earnings.

In June, Bloomberg reported that the company was considering a sale and had held talks with banks. Shares climbed 20% higher on the news.